KR20010052341A - Quantitative methods, systems and apparatuses for gene expression analysis - Google Patents

Abstract

The present invention provides a method of quantifying the association of a first gene expression profile with a second gene profile and aligning the association of a plurality of gene expression profiles with respect to a single preselected gene profile. This method is useful for quantifying the relevance of environmental conditions to cells, for example, the effects of pharmaceutical drugs on cells. This method is also useful for quantifying the relevance of preselected environmental conditions to defined gene mutations in cells, and for quantifying the relevance of multiple gene mutations. It also presents a system and apparatus for implementing this method. Furthermore, provided are quantitative methods, systems, and apparatus for selecting a beneficial subset of genes for gene expression analysis.

Description

Quantitative Methods, Systems, and Apparatus for Gene Expression Analysis {QUANTITATIVE METHODS, SYSTEMS AND APPARATUSES FOR GENE EXPRESSION ANALYSIS}

In traditional drug development, specific drug targets such as enzymes on known biochemical pathways are first selected. Thereafter, one or a plurality of in vitro or in vivo assays specific for the selected target should be developed. Only after a target is selected and specific assays have been developed can a chemical compound possessing the desired activity be selected. Once the compounds possessing the desired activity for the selected targets in the designated assays are identified, these early prototyping compounds play a role in the development of derivatives with more beneficial therapeutic, pharmacokinetic and clinical properties. The bioactivity of these derivatives is assessed using the same designated assays as when identifying the dominating compounds.

Each of the above mentioned steps in a traditional drug development prototype causes compounds that were prospective in preclinical testing to fail clinical trials.

First, the selection of drug targets presupposes knowledge of the biological pathways clinically associated with the disease or etiology that the drug targets. After the clinical trial begins, the selected target may prove to be physiologically inappropriate. For example, a target may be involved in a number of related or unrelated biopaths. Designated in vitro assays may not confirm the effects of candidate drugs on these parallel or intersecting biological pathways. As a result, if a drug that has a positive effect on target activity in vitro is administered in vivo, it may cause unexpected toxicity or undesirable side effects.

Second, in vitro assays are not sensitive enough or specific enough. This problem may be encountered when the same assay is used for the development of the prodrug derivatives.

Therefore, there is a continuing need for improved strategies for drug development in the pharmaceutical field. In particular, there is a need for a drug development plan that does not rely much on proper initial screening. There is also a need for a drug development strategy that avoids separation of selected targets from the biopath during preclinical drug development. There is also a need for drug development methods that identify novel pathways and biological pathways associated with a pathology or disease or disorder of interest.

Recent technological advances in the field of measuring gene expression have made it possible to simultaneously measure the expression of multiple genes transcribed in prokaryotic or eukaryotic cells. The ability to create such gene expression profiles has led to material materials that can be used to study novel drug development strategies (Ashby et al., United States Patent No. 5,549,588).

To date, most gene profiles have separated nucleic acid expression products from host cells, labeled products (e.g., fluorescent or radionuclide labels), and placed them on spatially-addressable matrices consisting of surface-fixed DNA units bearing individual sequences. Labeled nucleic acids were made by hybridization (Lashkari et al., Proc. Natl. Acad. Sci. USA, 94, pp. 13057-62 (1997); DeRisi et al., Science, 278, pp. 680-86 ( 1997); Wodicka et al., Nature Biotechnology, 15, pp. 1359-67 (1997); and Pietu et al., Genome Research, 6, pp. 492-503 (1986).

The components of the matrix are chosen to represent the entire genes that can be expressed by the host, wherein the fixed DNA matrix is made from the host. As recorded by scanning laser, scanning confocal fluorescence microscopy or imprinters, the specific hybridization with various DNA elements on the matrix suggests the expression of individual genes. The nature of the individual genes is encoded as the spatial location of the elements in the matrix. Data is acquired, digitized and stored electronically. The combined data identifies the nature of the subset of genes expressed by the selected cell culture.

Ashby et al., US Pat. 5,549,588 discloses other ways of creating gene expression profiles. Ashby presented a "genome reporter matrix" wherein each element of the spatially-addressable matrix consists of one or a plurality of identical cells (or clones of cells) rather than specific nucleic acid sequences. At each matrix position, the cell carries a recombinant construct that directs expression of the common reporter gene from separate transcriptional regulatory elements. Transcriptional regulatory elements can be derived from a number of eukaryotic or prokaryotic cells. Sufficient amounts of matrix elements and transcriptional regulatory elements are included to provide representative sampling of the gene expression repertoire of the selected microorganism.

In order to measure gene expression, Ashby et al. Read the matrix directly by scanning with a detection device suitable for the reporter and directed to it. In one embodiment, the reporter encodes a protein that produces a fluorescent signal, eg, a green fluorescent protein, and thus scans with a fluorescence detector; In one embodiment, the reporter encodes an enzyme that produces a detectable signal with the photometer, and thus scans with the photometer. The signal recorded by the scanner suggests the expression regulated by each transcriptional regulatory element, the nature of which is encoded in the spatial position of the element in the matrix.

The foregoing technical criteria for creating gene expression profiles (hereinafter collectively referred to as "expression matrices") provide a great deal of information about the simultaneous expression of genes in cells under defined conditions.

Those skilled in the art thus emphasized the qualitative comparison of gene expression profiles, eg the identification of subsets of genes showing modified levels of expression under different conditions; Those skilled in the art have highlighted data manipulation that is not possible for quantitative analysis of large scale multidimensional data (Ashby et al. (Supra); Lashkari et al. (Supra); DeRisi et al. (Supra); Rine et al., WO 98 / 06874; and Seilhamer et al., WO 95/20681 (each of which is incorporated herein by reference).

These qualitative methods do not allow iterative calculation of relevance of the entire gene profile. Thus, it is desirable to create quantitative gene expression profiles and use this information to quantitatively compare the relevance of gene expression in selected cells under various environmental conditions (eg, treatment with different compounds).

Therefore, there is a need for a method of quantifying the relationship between a first gene expression profile and a second gene expression profile. There is also a need for a method capable of ordering the relevance of multiple gene expression profiles to a single pre-selected gene expression profile. In addition, there is a need for a quantitative method and apparatus for retrieving stored data (ie, gene expression profile data obtained in previous experiments) and for performing new comparative analysis on relevance.

Recent technological advances in the field of gene expression measurement have made it possible to simultaneously measure the expression of multiple genes transcribed in prokaryotic or eukaryotic cells, but from a technical point of view only very few expressed genes have been analyzed. For example, if a sample of a drug candidate is produced in small quantities by combinatorial chemistry, the amount is limited; Few drugs can test the effects on all genes of any cell type. In addition, analyzing each candidate drug in the entire expressible gene of the cell is too costly.

These problems are encountered when the genome to be analyzed is complex. Thus, to evaluate the effect of drugs or other environmental factors on each of the expressible genes of yeast cells, such as Saccharomyces cerevisiae, the expression of approximately 6,000 genes should be measured; To perform similar analysis on gene expression of nematodes, such as nematodes (C. elegans), the expression of approximately 20,000 genes must be measured; To assess the effects of drugs or other environmental factors on the expressible genes of human cells, approximately 100,000 genes should be measured.

In addition, not all genes are equally useful. Some do not obtain sufficient information due to the lack of dynamic range of expression, regardless of environmental conditions. Different genes have a variety of expressions, and much information can be obtained. Thus, U.S. Patent No. 5,811,231 (Farr et al.,) And European Patent EP 0680517 B1, present a selection of "stress genes" for identifying and characterizing compounds that are toxic to cells.

However, this approach requires prior knowledge of gene function. In addition, preconceived notions of such controlled screening reduce the likelihood of identifying previously unanticipated relationships; In a method useful for identifying such unexpected relationships, for example the method presented in this article, such controlled selection is particularly undesirable.

Another way is to select a subset completely randomly in the hope that the selected subset will represent the whole. The problem is that this selected subset is not useful for describing cellular conditions under one or more environmental conditions.

Another approach is to select genes identified as common responses to preselected environmental conditions, not common functions (Whitney et al., Nat. Biotechnol., 16: 1329-33 (1998). Located in the middle of any scheme, the scheme retains some of the disadvantages of both schemes.

Thus, there is a need in the art for methods that can select useful subsets of genes for expression analysis.

Summary of the invention

The present invention solves these and other problems in the art by providing methods, systems, and apparatus for quantitative analysis of gene profiles. In experimental examples, this assay quantifies the relevance of various drug treatments and sorts them in order to identify chemicals that act on the same molecular targets as those affected by the reference drug; Identifying chemicals that act in the same physiological pathway as that of the reference drug; It can be seen that the mechanism of action of a chemical can be demonstrated compared to the reference drug (without prior identification of the molecular target of the reference drug or development of a designated assay). This analysis can be applied to comparison of other cell phenotypes, including those caused by different environmental conditions or genetic perturbations (eg mutations).

In a first aspect, the present invention provides a method for quantifying the relationship between a first gene expression profile and a second gene expression profile. The first method comprises (a) generating a first gene expression signal and a second gene expression signal for each gene that is common to the first gene expression profile and the second gene expression profile; (b) formulate a relative expression score for each pair of first gene expression signal and second gene expression signal; (c) calculating a composite score from the formulated relative expression scores of these pair-units, which composite score quantifies the relevance of the two gene expression profiles.

In another aspect, the present invention provides a second method for quantifying the relevance of a first gene expression profile to a second gene expression profile, said second method being particularly suitable for comparing gene expression profiles obtained under light conditions.

The second method comprises (a) generating a first gene expression signal and a second gene expression signal for each gene that is common to the first gene expression signal and the second gene expression profile; (b) performing linear regression on several pairs of first and second gene expression signals for commonly occurring genes, wherein the correlation coefficient of this regression quantifies the relevance of the two gene expression profiles. do.

In a third aspect, the present invention provides a method of ordering the relevance of a plurality of gene expression profiles to a single preselected gene expression profile, the method comprising: (a) expressing a plurality of genes with respect to a preselected gene expression profile; Quantifying the relevance of each profile in pairs; (b) sorting the pair unit-measured amounts in order. In a suitable embodiment of this aspect of the invention, the pairwise quantification of relevance is carried out according to one of the two methods newly presented herein.

In a series of embodiments of the foregoing methods, the present invention provides a method for quantifying the relationship between a first environmental condition and a second environmental condition for a cell, the method comprising: (a) a first environmental condition and a second environmental condition; Obtaining a gene expression profile from cells under or genetically identical cells under; (b) quantifying the first gene expression profile and the second gene expression profile. In suitable embodiments, the first and second environmental conditions include exposure to chemical compounds, such as pharmaceutical drugs.

The invention also provides a method of ordering the relevance of a plurality of environmental conditions to a single preselected environmental condition in a cell, the method comprising: (a) pre-determining a plurality of environmental conditions from a cell or genetically identical cells; Obtaining a gene expression profile for each of the environmental conditions; (b) quantifying the relevance of the plurality of gene expression profiles in pairs relative to the preselected gene expression profile; (c) ordering the paired unit-measured quantities in order. In suitable embodiments, the environmental conditions consist of exposure to chemical compounds.

In another set of embodiments, the present invention provides a method of quantifying the relevance of a preselected environmental condition to a defined gene mutation in a cell, the method comprising: (a) determining a first gene expression profile in a cell with a defined mutation; Obtaining a second gene expression profile in wild-type cells under preselected environmental conditions; (b) quantifying the association of the first gene expression profile with the second gene expression profile.

The present invention also provides a method of ordering the relevance of each of a plurality of environmental conditions to a defined mutation of a cell, the method comprising: (a) a group of first genes from wild-type cells under one of the plurality of environmental conditions; Obtaining an expression profile of a second gene expression profile from cells with defined mutations; (b) quantifying the association of each of the first gene profiles in pairs relative to the second gene expression profile; (c) ordering the paired unit-measured quantities in order. In a suitable embodiment, the environmental conditions are those exposed to chemical compounds, and the pair-unit quantification is carried out according to one of the two new methods presented in this article.

In another embodiment, the present invention provides a method for quantifying the relevance of a cell's first gene mutation to a cell's second gene mutation, wherein the method comprises: (a) a first gene from a cell bearing the first gene mutation. Obtaining a gene expression profile, a second gene expression profile carrying a second gene mutation; (b) quantifying the association of the first gene expression profile with the second gene expression profile. The present invention also provides a method of ordering the relevance of each of a plurality of gene mutations to a preselected gene mutation of a cell, wherein the method comprises a first set of first gene expression profiles from a cell bearing one of the plurality of mutations. Obtaining a second gene expression profile from a cell having a preselected mutation; (b) quantifying the association of each of the first gene expression profiles in pairs relative to the second gene expression profile; (c) ordering the paired unit-measured quantities in order.

In a suitable embodiment, the environmental condition is to expose the cells to a chemical compound, the cells are yeast cells, preferably Saccharomyces cerevisiae, and the gene expression profile is obtained from the genome reporter matrix. The method can be broadly applied to gene expression profiles obtained from any environmental conditions (eukaryotic and prokaryotic cells, including humans) and other forms of expression matrices.

In another aspect, the present invention provides a system (including a computer system) for implementing the aforementioned quantitative method.

Thus, in this aspect, the present invention provides a system for quantifying the relationship between a first gene expression profile and a second gene expression profile, wherein the system is (a) common in both the first gene expression profile and the second gene expression profile. Means for producing a first gene expression signal and a second gene expression signal for each gene represented by; (b) means for formulating a relative expression score for the first gene expression signal and the second gene expression signal; (c) means for calculating a composite score from a pair of relative expression scores, wherein the composite score serves to quantify the relevance of the two gene expression profiles.

In a related aspect, the present invention provides a system for quantifying the relevance of a first gene expression profile and a second gene expression profile, the system being in common in (a) a profile of a first gene expression profile and a second gene expression profile. Means for generating a first gene expression signal and a second gene expression signal for each gene presenting; (b) means for performing linear regression on several pairs of first and second gene expression signals for commonly occurring genes, wherein the correlation coefficient of this regression is related to the relationship between the two gene expression profiles. Quantify

In another related aspect, the present invention provides a system for ranking the relevance of a plurality of gene expression profiles to a single preselected gene expression profile, the system comprising: (a) a plurality of genes for a preselected gene expression profile; Means for quantifying the relevance of each expression profile in pairs; (b) means for sorting the paired unit-measured quantities in order.

The present invention also provides a computer system for quantifying the relevance of a first gene expression profile and a second gene expression profile, said computer system comprising a processor, such as a digital microprocessor, (a) a first gene expression profile and Generate a first gene expression signal and a second gene expression signal for each gene that appears in common in the second gene expression profile; (b) formulate a relative expression score for each pair of first and second gene expression signals; (c) is calculated to calculate a composite score from the relative expression score of a pair-unit, where the composite score quantifies the relevance of the two gene expression profiles.

Similarly, the present invention provides a computer system for quantifying the association of a first gene expression profile with a second gene expression profile, wherein the computer system consists of a processor such as a digital microprocessor, and (a) the first gene expression profile Generating a first gene expression signal and a second gene expression signal for each gene commonly present in the second gene expression profile; (b) is programmed to perform linear regression on several pairs of first and second gene expression signals for commonly occurring genes, where the correlation coefficient of this regression quantifies the relevance of the two gene expression profiles. .

The invention also provides a computer system for arranging the relevance of a plurality of gene expression profiles to a preselected single gene expression profile, the computer system comprising a processor such as a digital microprocessor, and (a) a preselected Quantifying the association of each of the plurality of gene expression profiles in pairs relative to the gene expression profile; (b) are programmed to sort these pair unit-measured quantities in order. An apparatus comprising a programmable digital computer, an input means, and an exhibition means is provided, which can execute the calculation method presented on the input expression data and present the quantitative result to the combined exhibition means.

In another aspect, the invention provides a computer readable storage medium for storing instructions, wherein when the instructions are executed by a computer, the computer quantifies the association of the first gene expression profile with the second gene expression profile. And sorting the relevance of a plurality of gene expression profiles to preselected gene expression profiles, each of the novel methods described in this article.

In another aspect, the present invention provides a computer readable storage medium having a data structure applied to the method of the present invention. In this aspect, the present invention provides a computer readable storage medium having a data structure configured to store data that quantitatively associates a first gene expression profile with a second gene expression profile, the data structure for each expression profile. Consisting of a tuner and a scalar, said scalar quantitatively correlating a first gene expression profile with a second expression profile. The invention also provides a computer readable storage medium having a data structure configured to store data that ranks the relevance of a plurality of gene expression profiles to a preselected single gene expression profile, the medium comprising (a) a scalar A ranked list of, each scalar quantifies the relevance of each of the plurality of gene expression profiles in pairs to a preselected gene expression profile; (b) a tuner that associates each scalar with a respective gene expression profile.

Recent technological advances in the field of gene expression measurement have made it possible to simultaneously measure the expression of multiple genes transcribed in prokaryotic or eukaryotic cells, but from a technical point of view only very few expressed genes have been analyzed. For example, if a sample of a drug candidate is produced in small quantities by combinatorial chemistry, the amount is limited; Few drugs can test the effects on all genes of any cell type. In addition, analyzing each candidate drug in the entire expressible gene of the cell is too expensive.

Thus, in another aspect, the present invention provides a method of selecting a beneficial subset of genes for expression analysis. The present invention provides a cell phenotype selection method, which comprises selecting about 20% expressable genes in a cell for expression analysis, wherein simultaneous expression of the selected genes sufficiently limits the cell phenotype, Allows cell phenotypes to be quantitatively associated with other cell phenotypes. In these methods, 20% expressable genes are selected from cells, more preferably 15% expressable genes, more preferably 10% expressable genes, optimally 5% expressible. Genes, most preferably 1% -5% expressable genes and finally 1-2% expressable genes are selected. We also present algebras, computer systems, networks that influence this screening, and other devices that affect this method.

In one embodiment, the methods of the invention consist of selecting genes that exhibit the maximum range of expression in each gene group to which expression is associated. In suitable embodiments, the selection is performed on a group of genes that are common to a plurality of gene expression profiles, and each range and correlation is calculated from expression data on the plurality of gene expression profiles.

In a related aspect, the present invention provides a system for selecting a beneficial subset of genes for expression analysis, which system comprises means for selecting genes of the maximum expression range in each gene group to which expression correlates. In suitable embodiments, the selection is carried out on genes that are common in a plurality of gene expression profiles, and each range and correlation is calculated from expression data on the plurality of gene expression profiles.

The present invention also provides a computer system for selecting a beneficial subset of genes for expression analysis, the computer system consisting of: to select genes of the maximum expression range in each gene group to which expression correlates. A programmed processor (eg, digital microprocessor); A computer readable storage medium for storing instructions, where the instructions are executed by a computer, the computer implements a method of selecting a beneficial subset of genes for expression analysis analysis, the method wherein each gene to which expression is associated Selecting genes in the group of maximum expression range; A computer readable storage medium having a data structure configured to store data for determining a beneficial subset of genes for expression analysis, the data structure consisting of a group of gene tuners and optionally including instructions for gene function.

The present invention is directed to bioinformatics methods that can be used for pharmaceutical drug development. In particular, the present invention relates to methods, systems, and apparatus for quantitative analysis, comparison, storage, and visual presentation of gene expression profiles. The invention also relates to quantification methods, systems, and apparatus for selecting a beneficial subset of genes for expression analysis.

These and other objects and advantages of the present invention will become apparent from the following detailed description taken in conjunction with the accompanying drawings.

FIG. 1 is an integrated process diagram illustrating the process of deriving a gene expression signal suitable for quantitative analysis of a gene expression profile from a signal initially obtained in a gene expression matrix, FIG. 1A illustrates initial signal processing, and FIG. The optional continuous correction according to the adjusted adjustments will be described.

FIG. 2 is a scatter plot of the gene expression signal processed according to FIG. 1, wherein the signal is in two chemotherapeutic drugs known to be closely related in structure and function: 50 μg / ml daunarubicin and 50 μg / ml doxorubicin Derived from one individually processed genomic reporter matrix (see Example 2).

Figure 3 shows a curve of gene expression signals derived from a matrix separately treated with two drugs of heterologous structure and function: 50 μg / ml doxorubicin and 0.08 μg / ml myconazole.

Figure 4 shows a curve of gene expression signals derived from a matrix separately treated with two drugs of similar function to heterologous structure: 9 μg / ml mycophenolic acid and 50 μg / ml daunarubicin.

FIG. 5 is a consistent process diagram illustrating a first process of converting individual gene expression signals made according to the process planned in FIG. 1 into quantitative values for ranking the relevance of gene expression profiles.

6 is a consistent process diagram illustrating a first process of converting individual gene expression signals made according to the process planned in FIG. 1 into quantitative values for ranking the relevance of gene expression profiles.

FIG. 7 is a scatter plot of the gene expression signal actually processed according to FIG. 1, which signal is derived from a genomic reporter matrix consisting of 1532 isolated gene expression reporters, said matrix being closely related in structure and function. FIG. Chemotherapy drugs in dogs: Individual treatment with one of 10 μg / ml lovastatin (X axis) and 20 μg / ml (mevastatin (Y axis)).

FIG. 8 is a scatter plot of gene expression signals from 96 gene subsets of the 1532 gene expression signals shown in FIG. 7, wherein the subsets are selected according to the algorithms plotted in FIGS. 9 and 10.

9 is a consistent process diagram that illustrates the first of two important steps in the algorithm for selecting a beneficial subset of genes for quantitative analysis of gene expression profiles.

Figure 10 depicts two iterations of the second of two important algorithmic steps in order to select a beneficial subset of genes for quantitative analysis of gene expression profiles.

In order to fully understand the present invention, the following detailed description is provided. In the specification, the following terms are used.

In this article, "gene expression matrix" refers to a device for obtaining data on the simultaneous expression of multiple genes (Lashkari et al., Proc. Natl. Acad. Sci. USA, 94, pp. 13057-62 (1997). DeRisi et al., Science, 278, pp. 680-86 (1997); Wodicka et al., Nature Biotechnology, 15, pp. 1359-67 (1997); Pietu et al., Genome Research, 6, pp 492-503 (1996); Ashby et al., US Patent No. 5,549,588. "Genome reporter matrix" means a gene expression matrix such as Ashby.

"Gene expression profile" means established data that is permanently or temporarily stored in an electronic device or other device, each element of which represents a level of simultaneous expression of a clear and identifiable open translational framework of a cell, typically a gene. Obtained in the expression matrix.

In a first aspect, the present invention provides a method for quantifying the relevance of a first gene expression profile to a second gene expression profile, the method comprising (a) a common expression in a first gene expression profile and a second gene expression profile. Generating a first gene expression signal and a second gene expression signal for each gene; (b) formulating a relative expression score for each pair of first and second gene expression signals, and (c) calculating a composite score from the relative expression scores of a pair of units, wherein the composite score is Quantify the relevance of two gene expression profiles.

The present invention provides a second method for quantifying the relevance of a first gene expression profile and a second gene expression profile, the method of which (a) each gene commonly present in the first gene expression signal and the second gene expression profile. Generate a first gene expression signal and a second gene expression signal for, respectively; (b) performing linear regression on several pairs of first and second gene expression signals for commonly occurring genes, wherein the correlation coefficient of this regression quantifies the relevance of the two gene expression profiles. do.

The present invention provides a method of ordering the relevance of a plurality of gene expression profiles to a single preselected gene expression profile, the method comprising: (a) relevance of each of the plurality of gene expression profiles to a preselected gene expression profile; Quantitate in pairs; (b) sorting the pair unit-measured amounts in order. In a suitable embodiment of this aspect of the invention, the pairwise quantification of relevance is carried out according to one of the two methods newly presented herein.

Each of these methods can be more easily understood with reference to the drawings in detail.

Generalization of Individual Gene Expression Signals from Initial Expression Data

FIG. 1 is an integrated process diagram illustrating the process of deriving a gene expression signal suitable for quantitative analysis of a gene expression profile from a signal initially obtained in a gene expression matrix, FIG. 1A illustrates initial signal processing, and FIG. The optional continuous correction according to the adjusted adjustments will be described.

The initial data acquisition step, bounded by box 116, can be performed sequentially or simultaneously; Digitization 101 may be performed by a signal acquisition device itself using an independent analog-to-digital converter, or may be performed by directly acquiring expression data in a digital form.

Each of the successive data manipulation steps (FIGS. 1A, 1B, 5, 6) can be accomplished with a programmable digital computer using techniques known in the computer sciences. Some of the steps can be accomplished by using analog circuitry known in the art. Steps may be performed on a single computing device, a series of computing devices, or distributed to parallel cross-multiplying devices as long as the temporary ranking of the steps is kept. The process can be carried out continuously or discontinuously, with intermediate values stored at the identified stage for continuous processing.

From the program side of the digital computer, the steps shown in Figures 1, 5, 6, 9, and 10 are coded in one of the high-level languages known in the art, such as FORTRAN, BASIC, Pascal, C, C +, C ++, Java. ^TM etc. are included, but it is not limited to these;

Alternatively, the steps shown in Figures 1, 5, 6, 9, 10 can be coded directly in assembly language. The different steps are subroutines, macros or other commercially available statistical analysis programs (JMP? (SAS Institute) or UNISTAT? Statistical Package (Unistat, Ltd.)) or mathematical function programs (Mathematica ^TM (Wolfram Research, Inc.) Achievements using other entities included in). The choice of programming language can be facilitated by those skilled in the art.

As shown in FIG. 1, expression data is first obtained as an initial expression signal of a form and manner suitable for a particular gene expression matrix (100); In the case of expression matrices such as Ashby, fluorescence data can be obtained with a scanning laser. Initial expression signals are acquired separately for each physical location (matrix element) of the expression matrix. These initial expression signals represent the expression level of each gene analyzed in the matrix under selected environmental conditions.

Initial base signals are generally obtained from one or more regulatory sites on the gene expression matrix. The nature of these basic controls depends on the physical matrix. For example, these matrices that specify hybridization of fluorescently or radiolabeled nucleic acids may be one or a plurality of positions on the matrix that do not contain any nucleic acid as a control or one on a matrix that contains nucleic acids that are not complementary to a known ORF. It includes appropriate quantities from a plurality of positions. Similarly, a matrix measuring expression from recombinant reporters on transformed cells (Ashby et a,) includes a matrix containing cells lacking the recombinant reporter construct as a control; A matrix containing cells containing a recombinant construct that is unable to express a reporter gene; The reporter construct includes an appropriate amount from one or more locations on the matrix that contains cells that contain but lack the necessary substrate.

Although the basic control elements are generally included in each matrix, the basic quantities can also refer to data values previously obtained from different physical matrices or previously stored from similar matrices. The choice of type and number of such controls is within the skill of the person skilled in the art.

In general, the initial expression signal and the initial basic signal obtained as a signal representing the intensity of the fluorescence are then digitized 101 and then electrically stored as the initial signal value and the initial basic value. These data are stored using any convenient table, matrix or spreadsheet form, referred to collectively as a gene expression profile. The data can be stored as volatile data, such as values on random access storage. Alternatively, the data can be stored more permanently in a magnetic, optical or magneto optical storage medium.

As will be appreciated, the initial signal values for each individual element of the expression matrix can be individually determined by location on the corresponding multidimensional data matrix, by attaching head information to each component of the data, or by other suitable methods known to those skilled in the art. Distinguish clearly. For example, the fluorescence intensity of a single physical matrix element can be represented by a single record consisting of multiple fields, tuners for experimental operation, etc. The one or more fields identify the physical source of the signal, data, and data acquisition time.

As will be appreciated, the dynamic range of the initial expression signal is defined by the physical constraints of the shape of the expression matrix, in particular the dynamic range of the expression reporter and the sensitivity range of the acquisition device. As you can see, analog signals can be represented as initial signal values as digital data with varying depths, such as 8-bit, 16-bit, 32-bit, etc. The deeper the data, the stronger the encoded strength is. On the other hand, the need for storing these data becomes even greater. The data depth is therefore selected based on empirical needs that one of ordinary skill in the art can understand. In addition, initial digitization is performed using one data depth, and subsequent analysis proceeds to data with reduced depth. In the latter case, simple linear transformations can be used to reduce the data depth.

In an appropriate manner, use floating point numbers.

Since the initial expression signal from a number of matrix locations is low (ie subbasal), this process is not necessary to perform a baseline correction 118 whenever possible. Several methods of making such corrections are known in the art. In one approach, the measured (previous) baseline value is added to each initial signal value regardless of the initial value entered. Alternatively, add ½ of the measured base value to each input value.

Although these known or other suitable methods can be used, the following methods are preferred. Each initial signal value is compared with the initial basic value (102). If the signal value is equal to or exceeds the basic value, no correction is performed, and the transition signal is designated as the initial signal value (106). Alternatively, if the initial signal value is less than the base value, the signal is designated as the base value (104).

This approach is more conservative than the previous approach to basic calibration. It is assumed that the first signal A is 0 and the second signal B is equal to the base level BKG. In the first manner in which BKG is added to each signal value, the A value is equal to the BKG, and the B value is twice the BKG, and therefore B is artificially set to be twice the A value. In a second scheme of adding ½ BKG to each signal value, A becomes ½ BKG and B becomes 1½ BKG, so that the B value is artificially set to be three times the A value. In an appropriate manner, adjust only A to BKG and keep B to BKG so that the B value after calibration is not greater than the A value.

Using this conservative approach to basic correction, the object of the present invention is to deepen the use of as many acquired gene expression signals as possible to produce a composite score that quantitatively correlates one or multiple gene expression profiles.

Prior methods generally assess changes in intracellular gene expression by recording changes in expression levels based on gene units. Even when multiple genes are measured simultaneously, changes are recorded as multidimensional data (Lashkari et al.,). When looking at changes in the expression of an arbitrary gene—even when looking at changes in the expression of a set of individual genes—signals should be used that are almost unchanged in this expression comparison due to the presence of measurement errors.

In general, changes in expression levels that do not exceed some of the multiple standard errors chosen are ignored. For example, less than 2 times, less than 5 times or even less than 10 times the expression changes on individual genes are generally ignored.

However, it is recognized in the present invention that many of these ignored data are actual changes in gene expression and, thus, are useful information for comparison of gene expression profiles. For example, Figures 2, 3, and 4 are scatter plots, with each scattered point showing the relative expression of individual genes under two identified conditions. The drawings are described in detail below.

For this purpose, it should be understood that the scales are logarithmic and that the dot markings on the horizontal and vertical axes of these figures are set at intervals of one natural log (e ¹ , e ² , e ^3, etc.). As can be seen, most of the data is in a square defined by the first dot in each direction on both axes. In other words, all the data in these squares were ignored in the analysis because they were below the natural logarithm (approximately 2.7 times) which were ignored because they were indistinguishable from the standard error. When ignoring changes less than two natural logs (e ² , or 7.4- times), all data on the square defined by the second point mark in each direction are removed from the analysis. As can be seen from the figure, most of the useful data is lost.

In the present invention, these data can be used. The significance of small changes in the expression of any gene cannot be measured due to the magnitude of the standard error, but the significance of the overall change can be measured in practice; While the prior art focuses on standard error in the measurement of significance, the present invention instead focuses on the mean standard error. On average, the overall change in gene expression between the two different environmental conditions is strongly correlated as shown below.

Therefore, it is preferable to use the retention retention correction for the base as described above in order to keep the plurality of data through the basic correction step.

The signal for each matrix element adjusted as far as possible relative to the base is then normalized to other identical experiments, ie control of data acquisition behavior in a single expression matrix or variation in individual data acquisition from multiple matrices.

The usefulness of standardizing expression signals has been recognized in the art even before high comparative measurements of expressed genes using a gene expression matrix are possible. Thus, individual gene expression values by Nozan blot analysis were normalized compared to expression values of continuous housekeeping genes probed simultaneously or sequentially in the same blot. In this way, the variability introduced by unequal gel loading, mRNA purity variation, and the like could be controlled.

The limitation of the previous approach was that the expression of the individual genes themselves selected for reference may be different. This problem is solved by measuring the total gene expression of a cell, e.g., the expression of "housekeeping gene" in the present invention, and by measuring the change in gene expression in the presence of the drug (unexpected effect in the prior literature). More complicated

There are several ways to normalize the signal to control for variation between experiments. One way is to consider the mean signal of all genes as constant, the other way is to normalize the root mean square of the signal, and the other way is to normalize the mean log of the signal values. Average log normalization effectively removes the outer signal, which is the farthest signal from the average signal value.

The proper method here is to assume that the mean signal in all genes is constant: normalization is achieved by dividing each signal by the sum of all signals as shown in 108 (FIG. 1A).

When only a small fraction of genes expressed in cells are evaluated, it is not necessary to assume that the mean gene expression signal should be constant. Thus, when selecting a small subset of genes—for the initial generation, subsequent quantitation or initial acquisition and subsequent analysis of gene expression profiles—the standardization step can optionally be omitted. Thus, the normalization step 108 was omitted from the analysis of 96 gene subsets during the quantitative analysis presented in Example 5 below: The normalization step was omitted because the assumption of constant mean expression was not valid.

As a final step 110 of making individual signal levels for quantitative gene expression profile analysis, the logarithm of each signal level is taken; In other words, the signal is specified as the logarithm of the signal value. You can also use log ₁₀ , but natural logs are preferred.

Performing comparative analysis using logarithms of signal values has three advantages. First, when converted to algebraic values, the equivalent change in expression level is evaluated equally regardless of whether it is a decrease or increase in expression.

For example, consider a 10-fold decrease and a 10-fold increase in the initial 1 value. A 10-fold decrease (0.1 units) is an absolute decrease of 0.9 units. A tenfold increase (10 units) is an absolute increase of 9.0 units. The absolute value of increase of 9.0 appears to be a much larger change in gene expression than the absolute decrease of value of 0.9. With log ₁₀ for each value, the three values, in contrast, are -1, 0, +1, and the decrease and increase are practically the same.

Computing algebraic values is another side benefit of being able to directly assess the nature of expression profile data. When comparing the two replication profiles, the calculated log ratio for all genes appears to be distributed near zero according to the standard variance from any measurement error. Standard statistical measurements can quantify the values reproducible in different experiments.

A third advantage of using algebraic values is that the division of values according to algebraic scales has an advantage in the visual display of data as shown in FIGS. 2-4 (see below).

The signal produced in the process of FIG. 1A ending in step 110 is suitable for use in quantitative analysis of gene expression profiles as illustrated in FIGS. 5 and 6. However, as shown in FIG. 1B, a series of additional steps is preferably performed.

Drugs are prepared in a variety of solvents, including organic solvents, which themselves have a variety of effects on gene expression. Thus, changes in the gene expression profile made by introducing a drug into a cell culture medium include (1) changes made by the drug and (2) changes caused by the solvent. The medium itself is one cause of the change as shown in Example 4 and Table 7. In addition, strains or cell-type differences exist between the cells analyzed.

Signals by solvent-matched, medium-matched, preferably strain-matched regulation, in order to modulate these environmental effects and thus focus on subsequent profile comparisons of changes in gene expression due to the action of the experimental drug. Must be deducted as detailed in FIG. 1B.

First, the corresponding control expression matrix initial expression signal and initial base signal are obtained. For example, as a control of the effects on gene expression profiles caused by the presence of methanol in the actinomycin D solution (Tables 1 and 2), other identical expression matrices (genome reporter matrices) were treated with only methanol at the same concentration, Initial expression signals and initial base signals are obtained therefrom.

Calibration for environmentally-corresponding controls is performed separately for each gene as shown in FIG. 1B.

First, the genetic signal from the corresponding control matrix (Signal _mc 132) is subtracted from the genetic signal 130 obtained from the experimental matrix (134).

Next, the artificial results introduced by the previous basic correction 118 prior to standardization are processed in the manner of two decision queries 136 and 140. Queries can be performed sequentially in any order, or more generally in a single line of code.

If is less than or equal to 0, the correction signal 134 - in other words, when the Signal _mc 132 exceeds the test signal 130 of --Signal _mc is increased artificially for basic correction of previous standards (104), Signal _mc The actual value may be less than or equal to the signal 130. Accordingly, the first decision query 136 asks whether the corrected signal 134 is equal to or less than zero and whether signal _mc is equal to or less than basic in step 102. If the first decision query 136 turns true, then the corrected signal is set to zero (138). In other words, since it is impossible to determine whether the corrected signal is true, set the value to zero so that the signal is removed from the continuous analysis.

Similarly, if the corrected signal 134 is greater than zero--that is, if the experimental signal 130 exceeds the corresponding control signal _mc 132--the test signal 130 prior to normalization is artificially calibrated during the basic calibration. , The actual value of the signal 130 may be less than or equal to signal _mc . Thus, if the second decision query 140 changes to true, the corrected signal is set to zero (142).

Figures 2, 3 and 4 show scatter plots of gene expression data processed as described above, including the steps shown in Figures 1A and 1B.

Data in Figures 2-4 are derived from the initial expression signal produced by the genome reporter matrix (see Examples below). Figure 2 shows the data derived from the matrix individually treated with one of two chemotherapeutic drugs known to be closely related in structure and function: daunarubicin and doxorubicin. Figure 3 shows the data from the matrix separately treated with one of two drugs with different structures and functions: doxorubicin (chemotherapy drug) and myconazole (antifungal agent). Figure 4 shows the data derived from the chemotherapy drugs with different structures but with associated functions: matrixes treated separately with either mycophenolic acid or daunarubicin that inhibit DNA synthesis.

Each point on the graph of Figures 2, 3, and 4 represents the expression of a particular gene: X is calculated from the signal obtained in the presence of one of the drugs (doxorubicin in Figure 2, doxorubicin in Figure 3, Daunarubicin in Figure 4). The numerical value is adjusted to the point, and Y adjusts the numerical value calculated from the signal obtained in the presence of the drug (Daunarubisin in FIG. 2, myconazole in FIG. 3, mycophenolic acid in FIG. 4) to the point.

In the visual examinations of Figures 2, 3, and 4, the usefulness of expression profile analysis can be seen to facilitate drug development, and in addition to the extreme relevance (or irrelevance) presented in these figures, Instant quantitative analysis is also useful.

For example, an instantaneous examination of FIG. 2 shows that the two drugs are not identical but similarly affect the expression of most yeast genes: Each gene with increased expression by daunarubicin is equally doxorubicin Increasing; Each gene whose expression was reduced by daunarubicin treatment is equally inhibited by doxorubicin treatment; Each gene whose expression is not affected by Daunarubicin treatment is similarly unaffected by doxorubicin treatment.

In contrast, similarly dotted data from gene expression profiles made using the irrelevant drug doxorubicin and myconazole show significantly different patterns (FIG. 3). As shown in the figure, the expression of some genes is increased by two drugs (dots in the upper right quadrant), the expression of some genes is decreased by treatment of the two drugs (dots in the lower left quadrant), and the expression of other genes. Is adversely affected by the drug (points on the upper left and lower right quadrants).

Figure 4 shows an intermediate case, where the two drugs are known to affect DNA synthesis although they differ in mechanism.

Thus, quantitative evaluation of drug relevance is possible. These drugs (or other environmental conditions) with a dispersed point distribution similar to that shown in Figure 2 are closely related in action; Drugs with a scattered point distribution similar to that of Figure 3 are functionally irrelevant; Drugs with a scattered point distribution similar to that of FIG. 4 have different but somewhat related mechanisms of action.

In the case of known potent compounds, derivatives and analogs can be selected by identifying drugs with similar activity without resorting to precise biochemical assays. In fact, the mechanism of action of the lead compound does not need to be known. However, the potential of this analysis is limited by the ability to recognize these association patterns. This problem, which is minimal in the extremes shown in FIGS. 2 and 3, is more apparent in the intermediate case as in FIG. 4. The present invention solves this problem by providing a reproducible quantitative assessment of gene expression profiles; In the present invention, since the analysis of two or more compounds is possible, the relevance of the gene expression profile produced can be arranged in order.

How to quantify the relevance of gene expression profiles by making composite scores

The present invention provides a method for quantifying the relationship between a first gene expression profile and a second gene expression profile, the method comprising: (a) for each gene in which the first gene expression propyloid is common to the second gene expression profile; Generate a first genetic signal and a second genetic signal relative to each other; (b) formulate a relative expression score for each pair of first and second gene expression signals; (c) calculating a composite score from the pair-unit relative expression scores, wherein the composite score quantifies the relevance of the two gene expression profiles.

The first step of the invention has been described above, with reference to Figures 1A and 1B. The second and third steps will be described with reference to FIG.

In summary, relative expression scores 524 are individually formulated for each gene that appears common in both gene expression profiles (528). The composite score is then calculated from the overall relative expression scores of these individual genes (526), where the composite score serves to quantify the relevance of the two gene expression profiles.

As described in detail in Fig. 5, signals for the genes under the first condition (Signal 1, 500) are input. This signal is processed as shown in Figure 1; As mentioned above, the signal is corrected as shown in FIG. 1B, preferably by subtracting the environment-corresponding conditions, but this is not necessary. Signals 2 and 502 for the same gene under the second condition are similarly processed and subtracted as shown in FIG. 1 to produce a relative expression score 504. Since the signal input value is logarithm 110, the difference is the rate of expression.

The artificial result 118 introduced by the basic correction before standardization should be processed at this point as described above after the subtraction of the corresponding control signal.

Artificial result correlation is performed using two decision queries (506 and 510). Queries can be performed sequentially in any order, or more generally in a single line of code.

If the relative expression score (score 504) is less than or equal to zero--that is, if signal 2 exceeds signal 1--signal 2 is artificially increased during the baseline correction prior to normalization (104), and the actual value of signal 2 is The signal may be 1 or less. Accordingly, the first decision query 506 asks whether the relative expression score 504 is less than or equal to zero and whether signal 2 is less than or equal to basic in step 102. If the first decision query 506 turns true, signal 2 is set to 0 (508). In other words, since it is impossible to determine whether the relative score is true, the value is set to zero so that the score is not included in the composite score 526.

Similarly, if the relative expression score 504 is greater than zero--that is, if signal 1 exceeds signal 2--signal 1 prior to normalization artificially increases (104) during base calibration, The actual number may be less than or equal to signal 2. Thus, if the second decision query 140 changes to true, the relative expression score is set to zero with the corrected signal (518) such that the relative score is not included in the composite score 526.

Next, a genetic unit entry criterion comparison is performed (522). Each expression matrix technique has its own entry criteria for detection, below which signals cannot be measured at confidence levels. For example, the oligonucleotide hybridization platform of Lashkari et al. Has different detection entry criteria than the cell genome reporter matrix of Ashby et al.

These entry criteria are empirically determined. In a simple manner, acquisition of a non-treated profile or acquisition of a profile from the same drug treated cells is performed twice. When comparing two replication profiles, the calculated log ratio for all genes appears to be near zero according to the mean distribution due to random measurement error (however, the signal-to-noise ratio must be adequate--signal Is low, the default correction distorts the distribution). The standard bias of this distribution provides a guide for establishing appropriate entry criteria.

Thus, if the absolute value of the relative expression score corrected for the basic anthropogenic result is less than or equal to the empirically set entry criterion 516, the score is assigned to 0 (518) and then not included in the composite score 526. . Currently, the appropriate entry criterion for data obtained from the genomic reporter matrix of Ashby et al. Is 0.7. Those skilled in the art can establish these empirical entry criteria using the statistical techniques described above. In addition, because the technology change or data acquisition technology is more effective than the existing data acquisition technology, this empirical entry criterion is likely to change. In Experimental Examples 1 to 4 below, an entry criterion of 1.0 was applied using previously collected data.

In addition, the step defined by the box 522 does not consider the direction of the gene expression change on the first gene profile and the second gene profile. It is of course necessary to set the score to 0 which does not exceed the user-defined entry criteria. For the remaining scores, it is desirable to remove the directionality by setting the absolute value 520 of any non-negative score as the score. 2 In determining the relevance of processing, the information content of gene suppression is treated equivalently to the information content of gene activation--use only the magnitude of the relative change.

Thus, it can be seen that there are two steps in the algorithm to set the relative expression score to zero and delete the data from the complex expression profile score. In steps 506, 508, 510, 512 defined by box 514, the score is set to zero if it is not known whether the direction of the relative score is true due to basic correction and standardization. In steps 516, 518, and 520 defined by box 522, the score is set to zero if not statistically indistinguishable from zero, although not an artificial result.

Based on the genetic units, the final operation 524 corrects the dynamic range of dynamics of gene expression clearly manifested by the various genes of the microorganism. For example, some genes can only change the gene expression by a factor of two regardless of the intensity of the change in condition; Other genes can change gene expression 200-fold. In order to prevent genes with a larger dynamic range from excessively distorting the comparative analysis, each of the relative expression scores is divided by the logarithm of the square root of the maximum expression observed for that gene in the previous experiment. As seen at 524, each of the relative expression scores is divided by the logarithm of the square root of the maximum signal output in step 108; In other words, each of the relative expression scores is divided by the logarithm of the square root of the largest normalized signal observed so far in the gene (½ log). As will be appreciated by those skilled in the art, the values for each gene will depend on the expression matrix technique (eg, array size) and the data collected, which can sometimes change with further experimentation.

An alternative is to consider the dynamic dynamic range of various genes at step 524.

In this alternative, each of the relative expression scores is divided by the logarithm of the square root of the maximum signal output thus far, i.e. the maximum normalized signal, of step 108, which differs from the first scheme in the values chosen to achieve normalization. ("Σ signal" in step 108). This approach is described in detail in Example 5 below.

In another embodiment, each of the TKDEOJR expression scores is divided by the logarithm of the square root of the maximum signal input in step 108; In other words, divide by the logarithm (½ log) of the square root of the maximal unnormalized signal for each of the genes of the relative expression score. This is particularly relevant where standardization is inadequate.

Alternatively, instead of dividing by the logarithm of the square root, one can divide by the magnitude of the maximum log signal—normalized or denormalized. In the present invention, the principle of selecting the maximum square root is that a certain type of error is converted into the square root of the signal. The correlation of the logarithm of the square root has been found to provide a more beneficial expression profile comparison.

Another approach is to make no corrections on the assumption that genes whose expression can change to maximum are more biologically important or at least more significant in assessing the relevance of environmental conditions.

Another alternative is to process different genes differently based on empirically-determined significance for the assay to be performed. For example, most genes were treated as described above by dividing by the logarithm of the square root of the maximum expression observed for that gene in a previous experiment. However, certain predetermined gene subsets are treated differently at this stage to increase or decrease their significance in subset analysis.

The above-described steps, which are entirely defined by box 528, are performed for each of the genes that are common in both the first and second gene expression profiles. For some expression matrices, such as measuring gene expression in prokaryotic or small eukaryotes (eg yeast), almost all open reading frameworks can be compared this way. For other platforms using mammalian cells, multiple genes are evaluated. Clearly, only genes that are commonly measured between the first and second environmental conditions can be used to create relative gene expression scores.

The final scalar value (referred to as a composite score) expressing the relevance of the gene expression profile of the two conditions in terms of the scalar value may be calculated as a sum (526). The smaller the resulting value, the more closely the gene expression profile under the two conditions compared, and the value is zero if they are exactly the same.

Although no additional correction is required, the sum is corrected for the percent of genes included in the score.

Useless genes, ie, the percentage of genes excluded at the step defined by box 514 by assigning their relative score to 0, affects the composite score. Thus, in optimal correction 526 for useless genes, the simple sum of the relative expression scores is multiplied by the ratio of the number of genes / useful genes.

The assays presented in Examples 1-4 below were performed on gene expression profiles obtained from matrices with 864 reporters. Although not shown in FIG. 5, the scores obtained in step 526 are selectively normalized to compare matrices of different sizes by expressing relative expression scores per 1000 genes. To achieve this standardization, the relative profile score 526 is multiplied by the ratio of total genes on the 1000 / matrix.

The method described above can quantify the sequence of relevance of two gene expression profiles: the smaller the composite score of the result, the greater the relevance of the profile; The greater the relevance of the profile, the greater the relevance of the overall gene expression status of the cell under the two different conditions from which the gene expression profile was obtained.

Thus, the relevance of two environmental conditions to the overall gene expression of a cell can be assessed quantitatively. Environmental conditions are cultures on different media as further revealed in Example 4 below. Alternatively, the two environmental conditions consist of treatment with two different chemicals, eg, pharmaceutical drug candidates, with the relevance of the drug gene profile suggesting the relevance of drug action as revealed by the composite score. This aspect of the invention is described in Examples 1-3.

The method can be used to quantitatively correlate preselected environmental conditions to defined genetic mutations in a cell, wherein the method expresses a first gene expression profile from a cell with a mutation and a second gene expression from wild-type cells under a preselected environmental condition. Obtaining a profile; (b) quantifying the relationship between the first gene expression profile and the second gene expression profile.

In a suitable embodiment of the invention, the environmental conditions for obtaining expression data from wild-type cells consist of exposing to selected chemical compounds. This approach, starting with limited mutations, allows quantitative identification of drug candidates that mimic gene mutations in effect. Conversely, starting with gene expression profiles of important pharmaceutical drugs, mutations that mimic the effects of drugs can be identified by quantifying the relationship between the profiles obtained in the presence of drugs and their gene expression profiles. As a result, the mechanism of drug action can be explained through the identification of all targets directly or indirectly affected by the drug. In addition, the relevance of the two mutations can be determined by quantitatively correlating gene expression profiles obtained from each additional drug.

When the quantitative method of the present invention is applied to the analysis of gene mutations, the cells are preferably yeast cells, more preferably Saccharomyces cerevisiae. Yeast is particularly well suited for this and other purposes of evaluating genetic mutations because (1) the full-length genome of S. cerevisiae has been sequenced, and (2) homologous recombination results in target deletion or insertion. It is easy to implement and (3) many basic metabolic pathways are preserved between yeast and humans (Lashkari et al). However, this method is always applicable when mutations are identified in cells of other prokaryotic or eukaryotic microbes.

While the foregoing description relates to a method for quantitatively associating a first gene expression profile with a second gene expression profile, the present invention also provides a method for ordering the relevance of a plurality of gene expression profiles in order.

In order to rank the relevance of multiple gene expression profiles in rank order, the relevance of each common index or reference profile is evaluated to obtain a series of composite scores. The composite scores are then sorted by rank, with lower scores suggesting a greater relevance for the exponential profile. This ranking is shown in the table below.

Accordingly, the present invention provides a method of ordering the relevance of environmental conditions to a single preselected environmental condition in a cell, the method comprising (a) gene expression for each of a plurality of environmental conditions from a cell or genetically identical cells. Obtaining the profile, the gene expression profile for the preselected environmental conditions; (b) quantifying the relevance of the plurality of gene expression profiles to preselected gene expression profiles in pairs; and (c) sorting the paired-unit measured quantities in order. In suitable embodiments, one or more environmental conditions consist of exposing the cell to a chemical compound.

Similarly, the present invention also provides a method of ordering the relevance of each of a plurality of environmental conditions in order to a defined genetic mutation of a cell, the method comprising: (a) wild-type cells under each one of the plurality of environmental conditions; Obtaining a first gene expression profile from the second gene expression profile from a cell carrying the defined mutation; (b) quantifying the association of each of the first gene expression profiles in pairs relative to the second gene expression profile; (c) consists of sorting the quantities measured in pair-units in order.

In a similar manner, the invention also provides a method of ordering the relevance of each of a plurality of gene mutations to a defined or preselected mutation of a cell, wherein the method comprises (a) each carrying one of the plurality of gene mutations. Obtaining a first gene expression profile from a cell and a second gene expression profile from a cell with a preselected mutation; (b) quantifying the association of each of the first gene expression profiles in pairs relative to the second gene expression profile; (c) consists of sorting the quantities measured in pair-units in order.

How to quantify the relevance of gene expression profiles by linear regression

The sequencing of the composite scores and relevance presented by the process of Figure 5 is significantly weighted by an external gene, ie a gene whose expression varies substantially between the two measured conditions. However, this also applies to the correction of the dynamic range of various gene expressions and to the results of steps 516, 518, 520 defined in box 522 in FIG. 5, where the measurements were made by applying the need for entry criteria for data entry. The expression in conditions reduces contribution by cells that are slightly modified. The advantage of this propensity is that it focuses only on the sequence of genes that most contribute to phenotypic changes.

Figure 6 provides an alternative method of quantitatively correlating gene expression profiles, which places priority on relevance in terms of commonality of the direction of change in individual gene expression, rather than magnitude of change. The method presented in FIG. 6 provides several advantages over the method presented in FIG. 5, in particular the ability to correlate gene expression profiles obtained using small concentrations of pharmaceutical drugs, which are currently mild treatments such as low concentration drugs. Used to quantify the relevance of the profile obtained under the conditions. Whether to select the algorithm shown in FIG. 5 or the algorithm shown in FIG. 6 is determined empirically after comparing the results. This choice is included in the knowledge of the art.

Before describing the details of this alternative method, the conceptual difference between the two methods can be optimally visualized by considering the scattered points of FIG. As noted above, FIG. 2 shows the scattered relative gene expression of different genes in yeast cells individually treated with two closely related anti-tumorigenic chemotherapy drugs. As mentioned above, the treatments appear to be closely related, each equally affecting the size and direction of individual gene expression: As a result, most points are roughly distributed over the line passing through the origin. As will be appreciated, the same conditions, basic members, noise members, and other variant members theoretically create a series of expression points, all of which are precisely located on the line passing through the origin.

The entry criteria applied in steps 516, 518, and 520 (limited to box 522) can be conceptualized as two equally sloped parallel lines that are equidistant from the regression line (some kind of confidence interval) inferred from the data in FIG. As the entry criterion applied empirically in step 516 is smaller, the nearest entry baseline can be considered to approach the data regression line; When the number of externally located data points increases, it is considered that the nearest entry reference line is far from the data regression line, and the number of externally present data decreases. Since only points outside the entry baseline contribute to the expression profile score (compare steps 518 and 520), the method presented in FIG. 5 is actually affected by the location where these points are away from the regression line.

In contrast, the method presented in FIG. 6 focuses on the degree to which the data points are fitted to a perfect regression line, meaning that they are theoretically identical in processing. Points located directly on the regression line without showing minimal significance in the analysis actually contribute to the score. This method focuses on the direction of change in gene expression instead of the magnitude of change in gene expression. This method is less sensitive to the concentrations of the various drug treatments to be compared, as shown in Example 3 below.

Figure 6 illustrates a second way of quantifying the relevance of two gene expression profiles.

As processed according to FIG. 1, the gene expression signal for each gene that is common to the first (signal 1, 600) gene expression profile and the second (signal 2, 601) gene expression profile is an input. The signal was further corrected for the corresponding control according to the number set in FIG. 1B.

Next, corrections are made for the dynamic range of dynamics of gene expression clearly revealed by the various genes of the microorganism in operations 610, 611-similar to those performed in step 524 with the previous logarithm set in FIG.

The same alternative of adjusting the dynamic range was described above in terms of step 524. Thus, signals 600 and 601 are divided by the logarithm of the square root of the maximum (normalized) signal output from step 108; Divide by the logarithm of the square root of the maximum (unnormalized) signal input in step 108; Dividing by the logarithm of the maximum signal—normalized or denormalized—not the logarithm of the square root; No change is made to correct dynamic range; Or it can be adjusted individually using empirically selected values. Another alternative illustrated in Figure 5 is to adjust the dynamic range of all genes to be analyzed by dividing by the logarithm of the square root of the maximum normalized value, where the values used for standardization are selected from a larger group of genes.

Next, the first expression signal (signals 1, 610) and the second expression signal (signals 2, 611) are correlated (620) to provide two-dimensional coordinates for each gene. Linear regression 625 in the set of paired data, representing the expression of all genes common in both gene expression profiles, provides a score 626 that provides a relevance quantitative measure of the two gene expression profiles. Implies a more closely related relationship. The correlation coefficient itself and any multiple of it can be used as the score. The scores presented in the following examples are obtained by multiplying the correlation coefficient by 100.

Thus, before the first algorithm (FIG. 5) sums the first expression signal and the second expression signal for each gene to obtain a composite score, both expression signals are combined into a single scalar value 504 (first gene expression profile and second). Indicative of the ratio of expression between gene expression profiles), the algorithm continues to retain values as individual coordinates until the final stage.

Any data structure can be used, such as a single two-dimensional matrix, a set of vectors, etc., such that the first and second signals for each of the genes that appear in common with respect to the linear regression purpose can be used. In addition, the proximity of the data fit to the optimal virtual line through the two-dimensional data can be used in accordance with the present invention for the calculation of the relative profile score in steps 625 and 626. One skilled in the art can identify these structures and statistical methods and encode these calculations in a digital computer; This closeness of Pitt found that the relevance of the newly revealed gene expression profile can be reliably and reproducibly quantified.

The steps identified in FIG. 6 can be added to the present invention.

Signals 1 601 and 2 602 are subject to the same questions as presented in 506 and 510. In other words, questions may arise whether previous basic correlations and averaging potentially hinder the definitive determination of the direction of expression change between the two conditions. In this case, in other words, if the query presented at 506 or 510 is true, the signal for the gene is selectively deleted from linear regression.

Using the method shown in FIG. 6 similar to that shown in FIG. 5, quantitatively assessing the relevance of the two environmental conditions to the overall cell expression of the cells; Quantitatively assessing the relevance of preselected environmental conditions to defined gene mutations in cells; The relevance of two different mutations can be quantified. In addition, the algorithms and methods shown in FIG. 6 can be used to rank the relevance of a plurality of gene profiles obtained from cells with various mutations, obtained from different environments, or obtained from a combination thereof. .

As described above, each gene that is common to the first gene expression profile and the second gene expression profile may be displayed on the gene expression profile regardless of whether the algorithm shown in FIG. 5 or the algorithm shown in FIG. 6 is applied. Same treatment as the gene. However, by weighting the expression changes of one or a plurality of preselected genes, one can increase or decrease their significance in the analysis. This weighting can be done by adjusting the signal in step 524 or steps 610 and 611.

Data storage

In each embodiment of the invention, irrespective of whether the method shown in FIG. 5 or the method shown in FIG. 6 is used, the data can be generated in any or all intermediate stages by individual gene expression profiles in the process shown in FIG. 1, 5 or 6. Can be stored. The data obtained from any single expression matrix is the digitized raw data obtained in step 101, the base-adjusted normalized signal obtained in step 108, the base-adjusted, standardized log signal obtained in step 110 or the corresponding obtained in step 112. It can be stored as a fully calibrated signal for the control.

A new comparison of relevance--in other words, a new calculation of a composite score according to the algorithm of Figure 5 or a relative profile according to the algorithm of Figure 6--can be performed using previously acquired and stored data. Therefore, if you perform additional experiments and install additional profile data obtained from the various gene expression matrices described in this article as described in this article, more data can be obtained for the above analysis. In particular, because more drugs are tested for their effect on overall gene expression, a much more comprehensive database can be prepared for comparison.

Each of the memories of the gene expression profile represents an individual cell state that is repeatedly referenced for comparison, where the memories edit the spectra that identify the individual states of the inactive material--NMR spectra, IR spectra, mass spectra, etc. Similar to what we have done, this comparison can identify unknown chemical structures. Conversely, the quantitative assessment of the relevance provided by the method and the apparatus described in this article can be adapted to other spectra, which modifications are known to those skilled in the art.

Drug Development and Other Uses for Quantitative Analysis of Gene Expression Profiles

The new methods of drug discovery have been derived from the quantification methods, systems and devices presented in this article. To quantify the relevance of a gene expression profile, a compound tests for similarity with a known mechanism of drug, a drug of known potency, or for similar mutations, conditions, diseases or disease states.

Regardless of whether the primary physiological process is perturbed by chemicals, treating the target cells with a drug changes the pattern of gene expression in the target cells. Similarly acting drugs cause similar patterns of change. The greater the similarity of action, the greater the similarity of changes in the gene expression profile. As a result, drugs that have a similar overall effect on intracellular gene expression with the ability to quantify the relevance of gene expression profiles; Inference, one can identify drugs with similar mechanisms of action;

If the mechanism of action of the first drug is known, identifying other chemical compounds that cause similar changes in the gene expression profile of the target cell is to identify additional compounds that share similar mechanisms of action. The mechanism of the first drug is unknown, but if the drug is known to be effective in treating any disease, identifying the drug causing a similar change in the gene expression profile of the target cell may be unknown, but not in the treatment of the pathology. Identify drugs that have a similar effect.

Thus, the ability to quantify the relevance of gene expression profiles eliminates the need to identify isolated pharmaceutical targets, develop directed assays, and select compounds for these activities in directed assays.

The ability to quantify the relevance of gene expression profiles also facilitates efforts during the later stages of drug development to narrow and focus on the specificity of action of anticipated drug candidates. For example, pharmacologically-effective derivatives of the protoplast compounds can be identified based on the quantitative relevance of the gene expression profile for derivatives of the protoplast candidate, as described above.

The following experimental examples illustrate some of the applications of the present quantitative methods.

In Example 1, the relevance of the drug to actinomycin D was evaluated by comparing the gene expression profile obtained in the presence of actinomycin D with a plurality of gene expression profiles obtained immediately after exposure to other pharmaceutical drugs. Using one of the algorithms described above, various concentrations of daunarubicin, 5-FUDR, doxorubicin, 5-FU, urea hydroxide, mycophenolic acid are quantitative for the overall gene expression of cells, especially S. cerevisiae. It was found to cause a similar effect. These drugs, such as actinomycin D, are known to affect nucleic acid synthesis.

Thus, if only actinomycin D is known, the data indicate that daunarubicin, doxorubicin, the nucleotide analogues 5-FUDR and 5-FU, and mycophenolic acid are similar drugs with known mechanisms of actinomycin D. It can be seen that. If we know that actinomycin D affects nucleic acid synthesis, daunarubicin, doxorubicin, nucleotide analogues 5-FUDR and 5-FU, mycophenolic acid in the data affect the nucleic acid synthesis and, thus, treat the treatment of cancer. It can be seen that they are useful as therapeutic drugs or to disrupt the life cycle of pathogens, in particular viral pathogens.

Conversely, if the mechanisms of all drugs except the reference are known, these data show that actinomycin D interferes with nucleic acid synthesis, from which the mechanisms can be insighted.

These insights do not require designated nucleic acid synthesis inhibition assays and prior confirmation of molecular targets for the drug. As a result, drugs with similar overall effects were identified, but no immobilized molecular targets were identified.

As measured by changes in overall gene expression in Examples 2 and 3, the relevance of multiple genes to one of two concentrations of daunarubicin was evaluated, where again, the relevance of the action was the structure of the preselected reference drug. Or it can be seen that the decision can be made without prior knowledge of the mechanism. It can be seen that in Example 4 the method presented in this article can be used more extensively to quantitatively correlate the effect on the cells of the overall environmental conditions.

How to Select Beneficial Gene Subsets for Gene Expression Profiling

The quantitatively compared gene expression profiles in the assays presented in Examples 1-4 each retain data on the coexpression levels of more than 800 different S. cerevisiae. These 800 genes are a subset of the microorganisms' expressable genes, estimated to be 6000. These results indicate that some of the cell's total gene expression needs to be analyzed to successfully apply the method presented in this article. Quantitative analysis provides information as a percentage of the estimated gene growth, but it is clear that only a few gene expressions will be used for this analysis.

Often, when technically considering the acquisition of gene expression data, it can be seen that only very few expressable genes are analyzed. For example, if a sample of a drug candidate is produced in small quantities by combinatorial chemistry, the amount is limited; Few drugs can test the effects on all genes of any cell type. In addition, analyzing each candidate drug in the entire expressible gene of the cell is too costly.

These problems are encountered when the genome to be analyzed is complex. Thus, to assess the effects of drugs or other environmental factors on each of the expressible genes of nematode genes such as C. elegans, the expression of approximately 20,000 genes should be measured; To assess the effects of drugs or other environmental factors on the expressible genes of human cells, approximately 100,000 genes should be measured.

In addition, not all genes are equally useful. Some do not obtain sufficient information due to the lack of dynamic range of expression, regardless of environmental conditions. Different genes have a variety of expressions, and much information can be obtained.

One way of selecting beneficial subsets of genes for expression analysis is to select genes individually by known or inferred functions. Thus, U.S. Patent No. 5,811,231 (Farr et al.,) And European Patent EP 0680517 B1, present a selection of "stress genes" for identifying and characterizing compounds that are toxic to cells.

However, this approach requires prior knowledge of gene function. In addition, preconceived notions of such controlled screening reduce the likelihood of identifying previously unanticipated relationships; In a method useful for identifying such unexpected relationships, for example the method presented in this article, such controlled selection is particularly undesirable.

Another way is to select a subset completely randomly in the hope that the selected subset will represent the whole. The problem is that this selected subset is not useful for describing cellular states under one or more environmental conditions.

Another approach is to select genes identified as common responses to preselected environmental conditions, not common functions (Whitney et al., Nat. Biotechnol., 16: 1329-33 (1998). Located in the middle of any scheme, the scheme retains some of the disadvantages of both schemes.

7 and 8 quantitatively show the results of a novel alternative to the selection of beneficial gene subsets for gene expression analysis described below. This novel approach involves the selection of genes for expression analysis based on diversity of expression—except size, direction or general.

FIG. 7 is a scatter plot of the gene expression signal processed according to FIG. 1, derived from a genomic reporter matrix consisting of 1532 individual S. cerevisiae gene expression reporters, each matrix closely in structure and function. Two drugs known to be involved are treated separately: one of 10 μg / ml lovastatin (X axis) and 20 μg / ml mevastatin (Y axis). As discussed above in FIG. 2, a review of the figures shows that the two drugs are not identical but similarly affect the expression of most yeast genes: a gene whose expression is increased by lovastatin, each gene is mevastatin. Increases equally by; Genes whose expression is reduced by lovastatin treatment, each gene is equally inhibited by mevastatin treatment; Genes whose expression is not affected by lovastatin treatment are similarly unaffected by mevastatin treatment. The result is that most data points lie on the line passing through the origin.

8 is a scatter plot of gene expression signals from 96 gene subsets selected from 1532 gene expression signals. Only one of the genes shown in FIG. 7 was selected for display in FIG. 8, but the strong correlation between the two drug treatments can also be seen here. A list of 96 genes in the selected subset is shown in Table 9 of Example 5 below. Genes on subsets selected without consideration of known functions appear to possess a variety of functions (the gene functions shown in the table are shown in the Stanford University Saccharomyces genome database (http: //genome-www.stanfor). .edu / Saccharomyce s)).

The gene subsets shown in FIG. 8 were selected from those shown in FIG. 7 during the process consisting of two basic algorithmic steps: In the first step, each of the genes shown in FIG. 7 was classified according to the maximum dynamic range of expression ; In the second step, the expression of almost all genes on the list is strongly correlated during the iteration. The result is that the response of the various genes seen in the disc persists in the selected subset, with each correlated gene group represented by a subset of genes with the largest dynamic range.

As illustrated in FIG. 8, it is a principle to select a subset of genes among a plurality of genes for which expression data has been acquired in advance, but this method has the greatest utility in acquiring a few beneficial gene expression signals in the gene expression matrix itself. see.

Examples 1-4 measure the expression of 864 out of 6000 genes that can potentially be expressed by S. cerevisiae, i.e., 14.4% of the total genes potentially expressable by cells, It is shown that the cell phenotype is quantitatively limited, and therefore the relevance of the cell state can be quantitatively determined. Example 5 selects a much smaller subset of potentially expressible genes—96 out of 6000 or 1.6% of potentially expressable genes—their expression quantitatively limits the cell phenotype, and therefore Is beneficial enough to quantitatively determine the relevance of

Accordingly, an important aspect of the present invention is to provide a cell phenotype selection method, which comprises selecting about 20% expressable genes in a cell for expression analysis, wherein simultaneous expression of the selected genes is a cell. The phenotype is sufficiently defined to allow the cell phenotype to be quantitatively associated with other cell phenotypes. In these methods, 20% expressable genes are selected from cells, more preferably 15% expressable genes, more preferably 10% expressable genes, optimally 5% expressible. Genes, most preferably 1% -5% expressable genes and finally 1-2% expressable genes are selected. We also present algebras, computer systems, networks that influence this screening, and other devices that affect this method.

Two basic steps in the algorithm for selecting a beneficial subset of expressible genes for expression analysis can be more readily understood with reference to FIGS. 9 and 10.

The first of two important steps in the algorithm is to arrange the genes according to the dynamic range of expression. Use historical data whenever possible: For each gene, the maximum and minimum values of signal 108 in the database of electronically stored gene expression profiles are determined according to a properly formulated query (or set of queries) 900. .

As mentioned above, gene expression data can be stored at all intermediate time points during the process identified in FIGS. 1, 5, 6. For the purpose of the algorithm step set in FIG. 9, the signal output in step 108 is used. If the signal output in step 108 does not exist in the database, the numerical value should be reconstructed from the stored numerical value-for example, if the signal value output in step 110 is stored, the signal output in step 108 reverses step 110, In other words, it can be calculated by exponentiation.

The expression range is calculated by the ratio of the maximum signal and the minimum signal (specified range = Signal _max / Signal _min ) (902). Other measures of dynamic range-"Signal _max -Signal _min " can be used, but the ratio is now appropriate.

Next, the entry criterion is applied by comparing the range obtained in step 902 with existing empirical figures (904). If the range exceeds the entry criteria, the gene is kept for later use; If the range does not exceed the entry criteria, the gene is excluded from further analysis. As shown in step 906, this exclusion can be easily accomplished by setting the range to zero. For the screening shown in FIG. 8 and illustrated in Example 5, 10 was set as the entry criterion. In other words, only those genes that showed more than 10-fold changes in gene expression levels across a set of factual gene expression profiles stored on a database persisted in the selected subset.

The selection of range entry criteria at this stage of the algorithm is determined by empirical needs, which is included in the common knowledge in the art. In general, a beneficial subset of moderately reduced size is provided at a 10-fold entry criterion.

However, you can set the entry criteria to about 1; In other words, the cutoff is completely removed. As a result, all other factors are constant and a larger subset of genes are selected. In addition, the entry criteria set at this stage need not be limited to the total number.

Therefore, the entry criteria is set to about 1 or preferably 1 or more. In general, the entry criteria is set to 2 or more, more preferably 3 or more, more preferably 4, 5, 6, 7, 8 or 9 or more, and most preferably 10 or more.

In addition, the entry criterion is set to 10 to 100, preferably about 50, more preferably about 25, most preferably 10-20.

Genes whose expression range exceeds the empirical entry criteria are then classified according to the expression range.

10 schematically illustrates repeating the second algorithm step twice from left to right. Shown on the left is a list of genes output at step 908, ranging from maximum to minimum dynamic range ranking. The genes excluded at step 906 are not visible due to inadequate dynamic range.

During the first iteration of the process, the first gene on the list serves as an index or reference gene. In turn, the genes in the list are taken consecutively to calculate the extent to which the expression of the gene correlates with the index gene throughout the entire stored gene expression profile. If the correlation (r ² ) exceeds the empirically established value, the gene is excluded from the set.

The effect of this step is to eliminate all genes whose expression strongly correlates with the index gene, “gene 1”; The high degree of correlation suggests that a large number of information provided by the expression of excluded genes is present in the information inherent in the expression levels of index genes. As shown at the bottom of FIG. 10, the exponential gene (“gene 1”) persists in the beneficial gene subset; As shown in the middle of Figure 10, genes that are highly correlated with them ("gene 3" and "gene 4" are excluded.) Because the list is arranged in order of maximum to minimum expression range, it persists in the associated group. A single gene is one that has the maximum dynamic range of expression.

During the second iteration of the process, the first gene after gene 1 (“gene 2” in FIG. 10) becomes the index or reference gene. This also persists, as shown at the bottom of the figure.

In turn, the genes in the list are taken consecutively to calculate the extent to which the expression of the gene correlates with the index gene throughout the entire stored gene expression profile. If the correlation (r ² ) exceeds the empirically established value, the gene is excluded from the set. The surviving (unrelated) gene ("gene 6") then becomes the index gene for the next iteration.

The process is repeated until the list is extinguished.

In iteratively carrying out the step of removing the gene whose expression correlates with the expression of the index gene, the correlation is preferably performed at the gene expression signal, which is the output of step 140 (ie, the output of box 141). The number of genes remaining in the final subset is determined by the total number of genes contributing data to the database of gene expression profiles, the range entry criteria applied in step 904, and the correlation entry criteria applied during the iterative process depicted in FIG. Two entry criteria values can be empirically adjusted to create a beneficial subset with a randomly selected number of genes.

Thus, in the analysis presented in Example 5 below, the range entry criterion and the correlation entry criterion were empirically adjusted to provide a subset with 96 genes equal to the number of wells of a standard microtiter plate. The correlation entry criterion was set at 0,675.

Once the subsets of the desired size are identified according to the algorithms shown in FIGS. 9 and 10, quantitative analysis can be performed using the subsets according to the algorithms shown in FIGS. 5 and 6. The analysis can be performed by selecting from a more comprehensive gene expression profile as in Example 5, or by obtaining the expected gene expression profile using the gene subsets identified in the reporter matrix.

Example 5 illustrates the description of 96 gene subsets in 1532 genes available in a database of stored gene expression profiles. Tables 8 and 10-Table 8 list the ranks of relevance using 1532 genes, and Table 10 lists the ranks of relevance of the same profile based on the 96 genes selected using the method described above. In comparison, it can be seen that the 96 gene subsets continue to have enough diversity to quantitatively arrange the relevance of gene expression profiles: HMG-CoA reductase inhibitors are closest to lovastatin in the data on both tables. It can be seen that the drug affecting other stages of the sterol biosynthetic pathway is the next drug closely related in effect.

Quantitative analysis of gene expression in Example 5 was performed on 96 gene subsets using the algorithm of FIG. 6 (FIGS. 1A, 1B, 6), and the algorithm of FIG. 5 (FIGS. 1A, 1B, 5) can also be used. have. In addition, a table of expression data for 96 genes obtained in Figure 8-index profile (rank 0) versus data obtained from profiles appearing in rank 2 (20 μg / ml mevastatin dissolved in 1% ethanol) is thus selected. The set also shows that it can be used for quantitation of gene expression profiles.

The following examples are for illustrative purposes only and do not limit the invention.

Relationship of the drug to 80 μg / ml actinomycin D

Replicated genomic reporter matrices were made according to Ashby et al. In short, for each of these matrix recombination constructs, these constructs that propel the fluorescent reporter from the distinct yeast promoter were individually transformed into separate cultures of Saccharomyces cerevisiae of the same strain background. Transformed cultures were selected to maintain reporters and to prevent contamination by untransformed cells. Each of these transformed yeast cultures was isolated and maintained in separate spatially-addressable wells of the matrix.

The matrix used has 864 isolated constructs, which allow simultaneous expression levels of more than 800 genes. Each matrix is placed under the limited environmental conditions specified in Tables 1 and 2. Each gene expression profile was obtained from each matrix presented by Ashby et al., Digitized and stored electronically.

Thereafter, the relationship between the profiles made in the presence of 80 μg / ml actinomycin D and each gene expression profile was determined according to the method shown in FIGS. 1A, 1B, 5 (Table 1) or 1A, 1B, 6 (Table 2). And quantified in pairs. The pairwise relevance figures are then arranged in rank order. The result is:

Tables 1 and 2 show that each method presented in this article can quantify the relevance of gene expression profiles, thus confirming the relevance of drug treatment.

Therefore, as shown in Table 1, it can be seen that the 60 μg / ml actinomycin D treatment in the algorithms of FIGS. 1A, 1B, 5 is the most closely related treatment for reference or exponential conditions, which is 80 Exposure to μg / ml actinomycin D. 40 μg / ml Actinomycin D followed by 50 μg / ml Actinomycin D.

Various concentrations of daunarubicin, 5-FUDR, doxorubicin, 5-FU, urea hydroxide, mycophenolic acid are next. Both of these drugs, similar to actinomycin D, affect nucleic acid synthesis. Treatment with anti-active drugs is much less relevant: treatment with yeast alpha factors is ranked 26 and 27, followed by mevastatin, the latter being an inhibitor of HMG-CoA reductase. Rank 31 is found in the profile made by treatment with environmentally-responsive controls, not drugs, followed by antifungal agents (myconazole and griseofulvin) treatment and calcium channel blocker (verapamil) treatment.

Thus, if only the mechanism of action of actinomycin D is known, the data shows that daunarubicin, 5-FUDR, doxorubicin, 5-FU, urea hydroxide, mycophenolic acid have a mechanism similar to the known mechanism of actinomycin D. It implies that it is a drug. If we know that actinomycin D interferes with nucleic acid synthesis, the data show that daunarubicin, 5-FUDR, doxorubicin, 5-FU, urea hydroxide, mycophenolic acid also affect nucleic acid synthesis and, thus, treatment of cancer To chemotherapy drugs or to disrupt the life cycle of pathogens, in particular viral pathogens.

Conversely, if the mechanisms of all of these drugs except the reference are known, these data show that actinomycin D interferes with nucleic acid synthesis, from which the mechanisms can be insighted.

These insights do not require designated nucleic acid synthesis inhibition assays and prior confirmation of molecular targets for the drug. As a result, drugs with similar overall effects were identified, but no immobilized molecular targets were identified.

Table 2 presents a quantitative ranking of the relevance of gene expression profiles made using the methods and algorithms of FIGS. 1A, 1B, 6, as applied to the same set of electronically-stored gene expression profile data.

As can be seen, the drugs affecting nucleic acid synthesis are located as closest to treatment with 80 μg / ml actinomycin D. Note that the arrangement of ranks is the arrangement of decreasing concentrations of actinomycin D.

Example 2

Relationship of drug to 50 μg / ml daunarubicin

Gene expression profiles were obtained and stored as shown in Example 1 and Ashby et al.

Thereafter, the relationship between the profiles made in the presence of 50 μg / ml daunarubicin and each gene expression profile was determined according to the method shown in FIGS. 1A, 1B, 5 (Table 3) or 1A, 1B, 6 (Table 4). Quantification was performed in pairs. The pairwise relevance figures are then arranged in rank order. The result is:

In the data presented in Table 3—using the method shown in FIG. 5—the drugs that work closely with daunarubicin have been identified as: doxorubicin, actinomycin D, 5-FU, 5-FUDR, which are these Consistent with the known activity of the drug. In contrast, in the data presented in Table 4—created using the method shown in FIG. 6—verapamil (calcium channel blocker), which appears to be closely related, is unclear.

Thus, in more severe treatments represented by high concentrations of the drug, the method shown in FIG. 5 appears to be more appropriate than the method shown in FIG. Example 3 below shows that the method presented in FIG. 6 is preferred for low concentrations of drug.

In the data of this example, it can be seen that the cloned gene expression profile, ie, the gene expression profile obtained in separate experiments under the same conditions, provides data of close ranking to each other, which shows the reproducibility of the assay.

Example 3

Relationship of the drug to 12.5 μg / ml daunarubicin

Gene expression profiles were obtained and stored as shown in Example 1 and Ashby et al.

Thereafter, the relationship between the profiles made in the presence of 12.5 μg / ml daunarubicin and each gene expression profile was determined according to the method shown in FIGS. 1A, 1B, 5 (Table 5) or 1A, 1B, 6 (Table 6). Quantification was performed in pairs. The pairwise relevance figures are then arranged in rank order. The result is:

The results presented in Tables 5 and 6 show the practical advantages of the second method of quantifying the relevance of gene expression profiles at low drug concentrations.

As shown in Table 5, the first method shown in FIG. 5 does not accurately quantify the association between the gene expression profile and the profile made in the presence of only 12.5 μg / ml daunarubicin, thus, 5% saline and 1000 μg. / Ml diltiazem (calcium channel blocker) is placed before 5-FU, which is placed prior to anaerobic growth and verapamil in rank.

In contrast, the same gene expression profile data analyzed according to the method shown in FIG. 6 (Table 6) newly aligns various concentrations of doxorubicin treatment with the most closely related 12.5 μg / ml daunarubicin treatment, which doxorubicin is It is known to be most closely related to daunarubicin in structure and function.

Example 4

Relevance of overall environmental conditions

The cloned genomic reporter matrix was made as presented by Example 1 and Ashby et al., Which holds 864 distinct elements that indicate the simultaneous expression of 84 different yeast open translational frameworks. Gene expression profiles were obtained, digitized and stored under the conditions shown below for each matrix. Then, the relationship between the profile produced by the cell culture in the yeast minimal medium and each gene expression profile was quantified in pairs, according to the method shown in Figure 1A, 1B, 5. The pairwise relevance figures are then arranged in rank order, and the results are shown in Table 7:

As shown in Table 7, the quantitative methods presented in this article can be arranged in order of relevance of the overall environmental conditions represented by changes in nutrient medium, as in separate treatment with individual drugs.

In addition, these data show that changes in media have a significant impact on overall gene expression, which shows the importance of correction for environmentally-corresponding controls, as shown in FIG. 1B.

Example 5

Selection of Beneficial Subsets of Genes for Quantitative Analysis of Gene Expression Profiles

Replicated genomic reporter matrices were made according to Ashby et al. The matrix used for the analysis presented in this example contains 1532 isolated constructs, thereby simultaneously expressing the expression levels of more than 1500 genes, ¼ of the genes that can be expressed by S. cerevisiae. It can be measured. Each matrix was subjected to limited environmental conditions as indicated in the individual descriptions in Tables 8 and 10, respectively. Gene expression profiles were obtained from each matrix, digitized and stored electronically as suggested by Ashby et al.

Thereafter, the relationship between the profile made in the presence of 10 μg / ml lovastatin and each gene expression profile is quantified in pairs according to the method shown in FIGS. 1A, 1B, and 6, which is similar to the previous method in two respects. Seems to make a difference.

First, standardization step 108 was omitted from the analysis of 96 gene subsets because the assumption of constant gene expression is not applicable to a small percentage of cellular genes.

Second, the correction of the reporter's dynamic dynamic range can be achieved in steps 610 and 611 by dividing each gene by the logarithm of the square root of the maximum normalized signal; In each case, however, the values used for standardization are appropriate for the 1532 gene subset.

Relevance figures in pairs are arranged in rank order. The result is:

Table 8 shows that applying the algorithms of Figures 1A, 1B and 6 to a gene expression profile with 1532 distinct gene reporters can quantify the relevance of the drug to 10 μg / ml lovastatin (HMG-CoA reductase inhibitor). -Consistent with the results presented in Examples 1-4 above.

Thus, other drugs of the same class-mevastatin, fluvastatin, simvastin, atorvastatin-appear to be most closely associated with lovastatin. Drugs that affect other stages of the sterol biosynthetic pathway, such as echonazol, clotrimazole, fluconazole, are placed after the ranks. Subsequently followed by drugs, progesterone, nifedipine, and tunicamycin, which exhibit substantially different structures or modes of action. Many other drugs with much lower relative profile scores are not shown.

The database of gene expression profiles used to create Figure 8 was then queried and applied to the algorithms depicted in Figures 9 and 10. This algorithm was designed to identify a subset of 1532 genes in a gene expression profile, which is a representative gene expression repertoire that can quantify the relevance of a gene expression profile regardless of the number of genes decreased. In order to achieve a subset of 96 genes-equal to the number of wells of a standard microtiter plate, the range entry criterion was empirically set to 10 and the correlation entry criterion was set to 0,675.

The gene subsets thus identified are shown in Table 9 below. The functions listed in the table for genes selected according to the present invention are the functions currently reported in the Saccharomyces genome database (http://genome-www.stanford.edu/Saccharomyces) of Stanford University.

As you can see, this subset, which is selected without consideration for gene function, includes a diverse set of genes that have invariant functions.

The relationship between the profiles made in the presence of 10 μg / ml lovastatin and each gene expression profile in the database was then determined according to the method shown in FIGS. 1A, 1B and 6 using only expression data obtained from the 96 genes shown in Table 9. Quantification in pairs. Relevance figures in pairs are arranged in rank order. The result is:

Table 10 shows that it is possible to select beneficial subsets of genes capable of quantitative analysis of gene expression profiles. The analysis shown in Table 10 using only 96 genes shown in Table 9, as in the analysis shown in Table 8 using data from all 1532 available genes, was the most closely related drug for lovastatin. CoA reductase drugs have been identified, with drugs acting on the same biosynthetic pathway appearing to be most closely related, and drugs that are completely independent of target and effect appear to be least relevant.

Although this demonstration was performed by selecting 96 genes from 1532 genes for which expression data were available from the database, through identification of this beneficial subset, gene expression data were then obtained from the identified reporters and thus obtained. The data can be used to quantify gene expression profiles.

All patents, publications, and other published materials mentioned in this article are hereby incorporated by reference in their entirety.

Having described the preferred embodiments of the present invention, it will be apparent to those skilled in the art that all changes and modifications without departing from the spirit and scope of the present invention will be included in the appended claims.

Claims (65)

In the method for quantifying the relationship between the first gene expression profile and the second gene expression profile, (a) generating a first gene expression signal and a second gene expression signal for each gene in common in the first gene expression profile and the first gene expression profile; (b) formulating relative expression scores for each pair of first and second gene expression signals; (c) calculating a composite score from the pair of relative expression scores, Wherein the composite score quantifies the relevance of two gene expression profiles. The method of claim 1, wherein the gene expression signal generation step comprises the following steps: (a1) comparing the magnitude of the initial expression signal obtained for each gene with the magnitude of the initial basic signal obtained for its relative gene expression profile; (a2) Adjust the magnitude of each of the initial expression signals smaller than the relative initial base signal. The method of claim 2, wherein the gene expression signal generation step (a3) further comprising normalizing the magnitude of the initial expression signal and the adjusted initial expression signal throughout the signal for the relative gene expression profile. The method of claim 3, wherein the gene expression signal generation step (a4) designating the logarithm of the normalized signal as a value for each gene expression signal. The method of claim 4, wherein the gene expression signal generation step (a5) subtracting from each standardized log signal, the same processed gene expression signal obtained for the gene from the condition-matched control. The method of claim 1, wherein the step of formulating a relative expression score consists of the following steps: (b1) for each of the first gene expression signal and the second gene expression signal, calculate a ratio between them; (b2) Each of the calculated ratios whose potential direction of the ratio has been changed by the basic signal adjustment and normalization steps is excluded from further processing. The method of claim 6 wherein the step of formulating a relative expression score further comprises the following steps: (b3) comparing the magnitude of the absolute value of the calculated ratio with the magnitude of the entry criterion constant; (b4) Each of the calculated ratios of absolute values not exceeding the entry criterion constant is excluded from further processing. 8. The method of claim 7, wherein formulating a relative expression score (b5) further comprising individually normalizing each of said expression scores with respect to the maximally observed maximum expression signal for the gene of said expression scores. The method of claim 6, wherein the step of formulating a relative expression score is (b6) separately normalizing each of said expression scores to the factually observed maximum expression signal for the gene of said expression scores. 10. The method of any one of claims 1-9, wherein the compound score calculation step consists of the following steps: (c1) accumulate all relative expression scores that were not previously removed; (c2) Adjust the percentage of genes previously removed. In the method for quantifying the relationship between the first gene expression profile and the second gene expression profile, (a) generating a first expression signal and a second expression signal for each gene common to the first gene expression profile and the second gene expression profile; (b) performing linear regression on the first and second gene expression signals in pairs for the commonly occurring genes, Wherein the correlation coefficient of the regression quantifies the relevance of the two gene profiles. The method of claim 11, wherein the gene expression signal generation step comprises the following steps: (a1) comparing the magnitude of the initial expression signal obtained for each gene with the magnitude of the initial basic signal obtained for its relative gene expression profile; (a2) Adjust the magnitude of each of the initial expression signals smaller than the relative initial base signal. The method of claim 12, wherein the gene expression signal generation step (a3) further comprising normalizing the magnitude of the initial expression signal and the adjusted initial expression signal throughout the signal for the relative gene expression profile. The method of claim 13, wherein the gene expression signal generation step (a4) designating the logarithm of the normalized signal as a value for each gene expression signal. The method of claim 14, wherein the gene expression signal generation step (a5) subtracting from each standardized log signal, the same processed gene expression signal obtained for the gene from the condition-matched control. 12. The method of claim 11, wherein the magnitude of the first gene expression signal and the second gene expression signal include signals of two or less natural logs. 17. The method of claim 16, wherein the magnitude of the first gene expression signal and the second gene expression signal include a signal of one natural log or less in size. A method of arranging the relevance of a plurality of gene expression profiles to a preselected single gene expression profile, (a) quantifying the association of each of the plurality of gene expression profiles in pairs relative to the preselected gene expression profile; (b) sorting the amounts measured in pairs in order of order. In a method for quantifying the relationship between a first environmental condition and a second environmental condition in a cell, (a) obtaining a gene expression profile from a cell or genetically identical cells under one of the first and second environmental conditions; (b) quantifying the association of the first gene expression profile with the second gene expression profile. 20. The method of claim 19, wherein quantifying the relevance of the gene expression profile is performed according to the method of any one of claims 1-9. 20. The method of claim 19, wherein quantifying the relevance of the gene expression profile is performed according to the method of any one of claims 11-17. 20. The method of claim 19, wherein the first and second environmental conditions consist of exposing the cell to a first chemical compound and a second chemical compound. A method of arranging the relevance of a plurality of genetic environmental conditions to a preselected environmental condition in a cell, (a) obtaining a gene expression profile for each of a preselected environmental condition and a plurality of environmental conditions from a cell or genetically identical cell; (b) quantifying the relevance of each of the plurality of gene expression profiles to the preselected gene expression profile; (c) sorting the amounts measured in pairs in order of order. 24. The method of claim 23, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 1. 24. The method of claim 23, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 11. The method of claim 23, wherein each environmental condition consists of exposing the cell to a chemical compound. A method of quantifying the relevance of preselected environmental conditions to defined gene mutations in cells, (a) obtaining a first gene expression profile from cells with defined mutations and a second gene expression profile from wild-type cells under preselected environmental conditions; (b) quantifying the association of the first gene expression profile with the second gene expression profile. The method of claim 27, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 1. The method of claim 27, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 11. 28. The method of claim 27, wherein the preselected environmental conditions consist of exposing the cell to a chemical compound. In a method of arranging the relevance of each of a plurality of environmental conditions in order to a limited gene mutation, (a) obtaining a first set of first gene expression profiles from subgenous cells under one of a plurality of environmental conditions, and a second gene expression profile from cells with defined mutations; (b) quantifying the association of each of the first gene profiles in pairs relative to the second gene expression file; (c) sorting the amounts measured in pairs in order of order. 32. The method of claim 31, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 1. 32. The method of claim 31, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 11. 32. The method of claim 31, wherein each environmental condition consists of exposing the cell to a chemical compound. A method for quantifying the relevance of a cell's first gene mutation to a cell's second gene mutation, (a) obtaining a first gene expression profile from a cell carrying a first gene mutation and a second gene expression profile from a cell carrying a second gene mutation; (b) quantifying the relationship between the first expression profile and the second expression profile. 36. The method of claim 35, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 1. 36. The method of claim 35, wherein quantifying the relevance of the gene expression profile is performed according to the method of claim 11. A method of sorting the relevance of each of a plurality of gene mutations in order to a preselected gene mutation in a cell, (a) obtaining a set of first gene expression profiles from each cell carrying one of the plurality of gene mutations and a second gene expression profile from cells carrying a preselected mutation; (b) quantifying the association of each of the first gene profiles in pairs relative to the second gene expression file; (c) sorting the amounts measured in pairs in order of order. A system for quantifying the relevance of a first gene expression profile to a second gene expression profile, wherein the system is (a) means for producing a first gene expression signal and a second gene expression signal for each gene that appears in common in the first gene expression profile and the second gene expression profile; (b) means for formulating a relative expression score for the first gene expression signal and the second gene expression signal; (c) means for calculating a composite score from a pair of relative expression scores, wherein the composite score quantifies the relevance of two gene expression profiles. A system for quantifying the relevance of a first gene expression profile to a second gene expression profile, wherein the system is (a) means for producing a first gene expression signal and a second gene expression signal for each gene that appears in common in the first gene expression profile and the second gene expression profile; (b) means for performing linear regression on the pair of first and second gene expression signals for commonly occurring genes, Wherein the correlation coefficient of the regression quantifies the relevance of the two gene expression profiles. A system for ranking the relevance of a plurality of gene expression profiles to a single preselected gene expression profile, the system comprising (a) means for quantifying the association of each of the plurality of gene expression profiles in pairs relative to the preselected gene expression profile; (b) means for sorting the paired unit-measured quantities in order. A computer system for quantifying the relevance of a first gene expression profile to a second gene expression profile, the computer system comprising (a) generating a first gene expression signal and a second gene expression signal for each gene that appears in common in the first gene expression profile and the second gene expression profile; (b) formulate a relative expression score for each pair of first and second gene expression signals; (c) a processor programmed to calculate a composite score from the relative expression score of a pair-unit, Wherein the composite score quantifies the relevance of two gene expression profiles. A computer system for quantifying the relevance of a first gene expression profile to a second gene expression profile, the computer system comprising (a) generating a first gene expression signal and a second gene expression signal for each gene that appears in common in the first gene expression profile and the second gene expression profile; (b) a processor programmed to perform linear regression on several pairs of first and second gene expression signals for commonly occurring genes, wherein the correlation coefficients of these regressions are determined by two gene expression profiles. A system characterized by quantifying relevance. A computer system for arranging the relevance of a plurality of gene expression profiles to a preselected single gene expression profile, the computer system (a) quantifying the association of each of the plurality of gene expression profiles in pairs relative to the preselected gene expression profile; (b) a processor configured to sort the paired unit-measured quantities in rank order. A computer readable storage medium storing instructions, wherein when the instructions are executed by a computer, the computer executes a method for quantifying the association of the first gene expression profile with the second gene expression profile. (a) generating a first gene expression signal and a second gene expression signal for each gene in common in the first gene expression profile and the first gene expression profile; (b) formulating relative expression scores for each pair of first and second gene expression signals; (c) calculating a composite score from the pair of relative expression scores, Wherein said composite score quantifies the relevance of two gene expression profiles. A computer readable storage medium storing instructions, wherein when the instructions are executed by a computer, the computer executes a method for quantifying the association of the first gene expression profile with the second gene expression profile. (a) generating a first expression signal and a second expression signal for each gene common to the first gene expression profile and the second gene expression profile; (b) performing linear regression on the first and second gene expression signals in pairs for the commonly occurring genes, Here, the correlation coefficient of the regression storage medium, characterized in that to quantify the relationship between the two gene profiles. A computer readable storage medium storing instructions, wherein when the instructions are executed by a computer, the computer executes a method for quantifying the association of the first gene expression profile with the second gene expression profile. (a) quantifying the association of each of the plurality of gene expression profiles in pairs relative to the preselected gene expression profile; (b) storage medium, characterized in that the ordered quantity measured in pairs. A computer readable storage medium having a data structure configured to store data quantitatively relating a first gene expression profile and a second gene expression profile, the data structure comprising a tuner and a scalar for each expression profile, wherein Scalar quantitatively correlates a first gene expression profile with a second expression profile. A computer readable storage medium having a data structure configured to store data in order of relevance of a plurality of gene expression profiles to a preselected single gene expression profile, (a) an ordered list of scalars, each scalar quantifying the association of each of a plurality of gene expression profiles to a preselected gene expression profile in pairs; (b) a storage medium comprising a tuner that associates each scalar with a respective gene expression profile. A method of selecting a beneficial subset of genes for expression analysis, characterized in that the gene having the maximum expression range is selected from each gene group to which expression is associated. 51. The method of claim 50, wherein said screening is performed on a group of genes that are common to a plurality of gene expression profiles. The method of claim 51, wherein each range and each correlation is calculated from expression data on a plurality of gene expression profiles. The method of claim 52, wherein the range is calculated as the ratio of maximum expression to minimum expression. The method of claim 52, wherein the screening step (a) arranging a group of genes that appear in common in a plurality of gene expression profiles, from highest to lowest in the range of expression; (b) selecting a gene having a maximum expression range in each gene group to which expression is associated in the plurality of gene expression profiles. 54. The method of claim 53, wherein the selecting step consists in successively repeating the following steps: (b1) selecting, for the first subset, the first gene remaining in the rank ordered set that has not yet been selected; (b2) calculating, from a plurality of gene expression profiles, a correlation of expression of each gene in said rank ordered set to expression of selected genes; (b3) Exclude all genes that show correlations above the entry criterion value from the rank ordered set. 54. The method of claim 53, wherein the ranking step further comprises a preceding step of removing all genes in a range less than the entry criterion value. A system for selecting a beneficial subset of genes for expression analysis, the system comprising means for selecting genes having the maximum expression range in each gene group with which expression is associated. A computer system for selecting a beneficial subset of genes for expression analysis, the computer system comprising a processor programmed to select a gene having a maximum expression range in each gene group with which expression is associated. A computer readable storage medium for storing instructions, wherein when the instructions are executed by a computer, the computer executes a method of selecting a beneficial subset of genes for expression analysis, the method maximizing in each group of genes with which expression is associated. A storage medium comprising the selection of a gene having a range of expression. A computer readable storage medium having a data structure configured to store data identifying beneficial subsets of genes for expression analysis, the computer readable storage medium comprising a set of gene tuners and optionally comprising instructions for gene function Storage medium. In the cell phenotypic selection method, the method consists in selecting about 20% of the expressable genes of the cell for expression analysis, Wherein co-expression of the selected genes sufficiently limits the phenotype of the cell, such that the phenotype is quantitatively associated with the phenotype of other cells. 62. The method of claim 61, wherein about 10% of the expressable genes of the cell are selected. 63. The method of claim 62, wherein about 5% of the expressible genes of the cell are selected. 64. The method of claim 63, wherein about 2% of the expressable genes of the cell are selected. 66. The method of claim 65, wherein about 1% of the expressible genes of the cells are selected.