JP2007535305A - Methods for molecular toxicity modeling - Google Patents

Abstract

The present invention is based on a method for constructing a toxicity prediction model using an algorithm for analyzing quantitative gene expression information and a method for predicting toxicity of a test agent. The present invention also includes a computer system that includes its toxicity prediction model along with a method of using the computer system by a remote user to determine the toxicity of a test agent.
[Selection figure] None

Description

Detailed Description of the Invention

(Related application)
[0001] This application claims the benefit of US provisional application 60 / 554,981, filed March 22, 2004, US provisional application 60 / 613,831 filed September 29, 2004; Both of these US provisional applications are hereby incorporated by reference in their entirety for all purposes. This application also claims the priority of international application PCT / US03 / 37556 filed on November 24, 2003, which is hereby incorporated by reference in its entirety for all purposes. .

(Submission of sequence listing on compact disc)
[0002] The compact disk sequence file submitted at the same time as this specification is based on US Patent Law Enforcement Rules Sections 1.821 (c) and 1.821 (e), and is incorporated herein by reference in its entirety. Incorporated in the specification. A four-part sequence listing is provided in which one part of each of the four compact discs is included. Copy 1, copy 2 and copy 3 are the same. Copy 1, copy 2 and copy 3 are the same for the CRF. Each electronic copy of the Sequence Listing was created on November 22, 2004 and has a file size of 2398 KB. The file names are as follows: copy 1-gene logic 5133-wo. txt; copy 2-gene logicg 5133-wo. txt; copy 3-gene logic 5133-wo. txt; and CRF-gene logic 5133-wo. txt.

(Background of the Invention)
[0003] The need for methods to assess the toxic effects of compounds, drugs or environmental pollutants on cells or organisms has led to the development of techniques that utilize organisms as biological monitors. The simplest and most convenient of these strains utilize unicellular microorganisms such as yeast and bacteria because they are most easily maintained and manipulated. In addition, single cell screening systems often use easily detectable changes in phenotype to monitor the effects of test compounds on cells. However, unicellular organisms are not an appropriate model for assessing the potential effects of many compounds on complex multicellular animals because they do not have the ability to perform biotransformation.

  [0004] Biotransformation of compounds by multicellular organisms is an important factor in determining the overall toxicity of drugs to which they are exposed. Accordingly, multicellular screening systems or screening systems using isolated eukaryotic cells are preferred or required to detect the toxic effects of the compounds. However, the use of multicellular organisms as toxicology screening tools has been significantly hampered by the lack of convenient screening mechanisms or endpoints, such as are available in yeast or bacterial systems. Furthermore, previous attempts to create toxicological prediction systems have failed to provide the necessary modeling and statistical information to accurately predict toxic responses (eg, WO00 / 12760, WO00 / 47761, WO00 / 63435, WO01 / 32928 and WO01 / 38579).

  [0005] The pharmaceutical industry spends significant resources to ensure that the subject therapeutic compounds are not toxic to humans. This process is long-term and expensive and requires testing in a range of organisms ranging from rats to dogs or non-human primates. In addition, modeling methods that design candidate drugs and their synthesis in nucleic acid, peptide, or organic compound libraries have increased the need for inexpensive, rapid, and accurate methods for predicting toxic reactions. Toxicity modeling methods based on nucleic acid hybridization platforms have made possible the use of biological samples obtained from animals or cell culture samples such as rat or rat hepatocyte cultures exposed to the compound, and to date Human organ toxicity should be detected much earlier than that.

(Summary of Invention)
[0006] The present invention compares to unexposed tissue or cells in animal tissues or cells, such as liver or kidney tissue or cells, which have been exposed to known toxins, particularly liver or kidney toxins. Based on the elucidation of global changes in all gene expression and the identification of individual genes that differ in expression by exposure to poisons.

  [0007] In various embodiments, the present invention provides gene expression information from samples exposed to drugs, samples exposed to toxins and control samples (samples exposed to vehicle, or non-toxic compounds or low levels). A method for predicting at least one toxic effect of a test agent by comparison with a database of gene expression information from samples exposed to toxic compounds of Such methods provide or generate quantitative gene expression information from the sample, and the RMA (robust multi-array average robust multi-array average) algorithm converts this gene expression information to a fold-change value. Transforming into a matrix, generating a gene regulatory score for each gene that is differentially expressed upon exposure to the test agent by means of a PLS (partial least square partial least squares) algorithm, and relating to the test agent Calculating a sample prediction score. This sample prediction score is then compared to a reference prediction score for one or more toxicity models. If the sample prediction score is equal to or greater than the reference prediction score, the test agent has at least one toxic effect or corresponds to a toxicity model that is compared to the prediction score of the test agent It can be predicted to cause at least one disease state.

  [0008] In various aspects, the invention includes a method of creating a toxicological model. Such methods provide or generate quantitative nucleic acid hybridization data for a plurality of genes in at least one cell or tissue sample exposed to a toxin and at least one cell or tissue sample exposed to the toxin vehicle. Transforming this hybridization data from at least one gene into a gene expression measure, such as the rate of change, by an RMA (Robust Multiarray Average) algorithm, by a PLS (Partial Least Squares) algorithm Generating a gene regulation score from a gene expression measure for at least one gene, and generating a toxicity reference prediction score for the toxin, thereby producing a toxicological module Including that to create a Le.

  [0009] In another aspect, the invention provides a computer-readable medium that includes a toxicity model for predicting the toxicity of a test agent, and a user provides a sample prediction score for the test agent and a toxicity reference prediction for the toxicity model. And a computer system that includes software that enables the prediction of at least one toxic effect of the test agent by comparison with the score.

  [0010] In a further aspect of the invention, gene expression information from tissues or cells exposed to the test agent may be prepared as a text file or binary file, such as a CEL file, and analyzed and on a remote central server. It may be transmitted over the internet for comparison with stored toxicity models. After processing, the user who sent the text file receives a report indicating the toxicity or non-toxicity of the test drug.

  [0011] In other aspects of the invention, a user may download one or more toxicity models to a local server from a remote central server and software for manipulating user data and toxicity models. . The gene expression information from tissues or cells exposed to the test agent may then be prepared as a text file such as a CEL file, and this information is analyzed and the toxicity model and user site stored on a local server. You can compare with. After processing, the software generates a report indicating the toxicity or non-toxicity of the test drug.

(table)
[0012] Table 1: Table 1 provides GLGC identifiers (fragment names in Table 2) for SEQ ID NOS and GenBank accession numbers for each gene fragment listed in Table 2, all of which are hereby incorporated by reference. And also incorporated by replication into the sequence listing attached hereto). Include gene name and Unigene cluster name.

  [0013] Table 2: Table 2 presents PLS scores (weighted gene index scores) from an exemplary general toxicity model of the kidney.

(Detailed explanation)
Definitions [0014] As used herein, "nucleic acid hybridization data" refers to any data derived by hybridization of a sample of nucleic acid with one or more series of reference nucleic acids. Such a reference nucleic acid may be in the form of a probe or a set of beads on a microarray, or in the form of a primer used in a polymerization reaction, such as PCR amplification, to detect primer hybridization with a sample nucleic acid. . Nucleic acid hybridization data may be in the form of numerical representations of hybridization and may be derived by techniques of quantitative, semi-quantitative, or non-quantitative analysis, or technology platforms. Nucleic acid hybridization data includes, but is not limited to, gene expression data. This data can be in any form, including fluorescence data or fluorescence probe intensity measurements from a microarray or other hybridization technology platform. Nucleic acid hybridization data can be raw data, or this data can be normalized to produce backgrounds caused by microarray high / low intensity spots, scratches, or high local or global background, as well as scanner electrical noise and Background or raw noise values including raw noise caused by sample quality fluctuations may be corrected or taken into account.

  [0015] As used herein, "cell or tissue sample" refers to one or more samples containing cells or tissues of animals or other organisms, including laboratory animals such as rats, mice and the like. A cell or tissue sample can comprise a mixed population of cells or tissues, or can be substantially a single cell or tissue type, such as hepatocytes or liver tissue. When used in the present invention, the cell or tissue sample may be a cell or tissue grown in vitro, such as a primary cell culture, an immortalized cell culture, a cultured hepatocyte, or a cultured liver tissue. The cell or tissue may be derived from any organ including, but not limited to, liver, kidney, heart, muscle (skeletal or cardiac muscle), or brain.

  [0016] As used in the present invention, "test agent" refers to an agent, compound, or composition that is being tested or analyzed by the methods of the invention. For example, the test drug can be a drug candidate for which toxicological data is desired.

  [0017] As used herein, a "test drug vehicle" refers to a diluent or carrier in which a test drug is dissolved, suspended, or administered to an animal, organism, or cell.

  [0018] As used herein, "toxin vehicle" refers to a diluent or carrier in which a toxin is dissolved, suspended, or administered to an animal, organism, or cell.

  [0019] As used herein, "gene expression measure" refers to any numerical representation of the expression level of a gene or gene fragment in a cell or tissue sample. “Gene expression measure” includes, but is not limited to, rate of change.

  [0020] As used herein, "at least one gene" refers to a nucleic acid molecule that is detected in a sample by a method of the invention. As used herein, the term “gene” refers to any fully characterized open reading frame and encoded mRNA and any cell or tissue sample assayed as described herein. RNA fragments that are expressed that can be detected by hybridization methods are included. For example, “gene” includes any type of nucleic acid that can be detected by hybridization with probes in a microarray, such as “gene” in Table 1. As used in the present invention, the at least one gene includes “a plurality of genes”.

  [0021] As used in the present invention, "fold change" refers to a gene, multiple, between experimental paradigms, such as a cell or tissue sample to be tested or treated compared to any reference or control. The numerical expression of the expression level of the gene or gene fragment. For example, by microarray for a gene or genes of a test cell or tissue sample compared to a control such as an unexposed cell or tissue sample or a cell or tissue sample exposed to a vehicle. It may be expressed as derived fluorescence or probe intensity. The RMA change factor described herein is a non-limiting example of the change factor calculated by the method of the present invention.

  [0022] As used in the present invention, a "gene regulatory score" is a weighted index score or a PLS score for each gene and a single gene derived from the rate of change from a treated sample relative to a control sample or Refers to a quantitative measure of gene expression for a gene fragment.

  [0023] As used in the present invention, a “sample prediction score” refers to a numerical score generated by the method of the invention as described herein. For example, using the PLS weight or PLS score for at least one gene in a gene expression profile generated from a sample, and the RMA change rate for that same gene, a “sample prediction score” can be calculated Good. By summing the individual gene regulation scores calculated for a given sample, a “sample prediction score” is derived.

  [0024] As used in the present invention, a “toxicity reference prediction score” refers to a numerical score generated from a toxicity model, which is a cutoff for predicting at least one toxic effect of a test agent. Can be used as a score. For example, the sample prediction score can be compared to a toxicity reference prediction score to determine whether the sample score exceeds or falls below the toxicity reference prediction score. A sample prediction score below the toxicity reference prediction score value is scored as not exhibiting at least one toxic effect, and a sample prediction score exceeding the toxicity reference prediction score value indicates at least one toxic effect. Scored as stuff.

  [0025] As used in the present invention, a logarithmic scale linear additive model includes any logarithmic linear model such as a logarithmic scale robust multi-array average, ie, RMA (Irzarry et al., Nucleic Acids Research 31 (4). e15 (2003)).

  [0026] When used in the present invention, "remote connection" refers to connection to the server by means other than direct wired connection. This term includes, but is not limited to, connecting to a server via a dial-up line, a broadband connection, a Wi-Fi connection, i.e., via the Internet.

  [0027] As used in the present invention, a "CEL file" refers to a file that contains average probe intensities associated with coordinate positions, cells, or features in a microarray (such information is provided by a CDF or ILQ file). ). See Affymetrix GeneChip® Expression Analysis Technical Manual herein.

  [0028] As used herein, a "gene expression profile" includes any quantitative presentation of expression of at least one mRNA species in a sample or population of cells, including differential display, PCR Profiles generated by various methods such as microarrays, other hybridization analyses, and the like.

Method for Creating Toxicity Models [0029] The present invention uses tests using selected compounds with well-characterized toxicity to assess and identify gene expression changes for which toxicity is predicted. A model or database may be built. The method of the present invention includes an RMA / PLS method that creates a model and database for predicting toxicity (analysis of raw gene expression data with a robust multi-array average algorithm with an assessment of predictive ability with a partial least squares algorithm) Is included.

  [0030] Generally, cell and tissue samples are analyzed after exposure to compounds known to exhibit at least one toxic effect. A low dose of such compound, or the vehicle in which the compound is prepared, is used as a negative control. Compounds known to show no at least one toxic effect are also used as negative controls.

  [0031] In the present invention, a toxicity test or "tox study" includes a set of cell or tissue samples that have been exposed to one or more toxins, including a toxin vehicle or a non-toxic, low-dose toxin. An exposed corresponding sample may be included. As described above, the cell or tissue sample may be exposed to toxin and control treatments in vivo or in vitro. In one test, exposure of a cell or tissue sample with toxins and controls can be performed by administering an appropriate dose to an animal model such as a laboratory rat. In some tests, exposure to a cell or tissue sample with toxins and controls can be performed by administering an appropriate dose to a sample of cells or tissue grown in vitro, such as primary rat or human hepatocytes. Such samples are usually organized in a cohort depending on the test compound, time (eg, the time from the start of administration of the test compound to the time the rat is sacrificed), and dose (amount of test compound administered). All cohorts in a tox test usually share the same vehicle control. For example, one cohort may be a set of samples from rats treated with acyclovir at a high dose (100 mg / kg) for 6 hours. A time-matched vehicle cohort is a set of samples that serve as a control for treated animals within the tox test, eg, a time-matched vehicle cohort for a high-dose sample treated for 6 hours with acyclovir And 6 hours vehicle treated sample.

  [0032] A toxicity database or "tox database" is a set of tox studies that include a reference database alone or in combination. For example, a reference database can include data from rat tissue and cell samples from rats treated with various doses of various test compounds and exposed to the test compounds for various lengths of time.

  [0033] RMA or robust multi-array average is the raw fluorescence intensity, such as that derived by hybridization of sample nucleic acid with an Affymetrix GeneChip® microarray, to the expression value, ie 1 for each gene fragment on the chip. (Irizarry et al. (2003), Nucleic Acids Res. 31 (4): e15, 8 pp .; Irizarry et al. (2003) "Exploration, normalization, and summers of high density. ): 249-264). In RMA, values typically occur between 4 and 12 on the log2 scale for genes expressed significantly above or below control levels. Such an RMA value can be positive or negative, and is centered around 0 when the rate of change is approximately 1. The matrix of gene expression values generated by RMA can be subjected to PLS to create a model for predicting toxic response, for example, a model for predicting liver or kidney toxicity. In a preferred embodiment, this model is verified by techniques known to those skilled in the art. It is preferred to use a cross-validation technique. In such a technique, the model success rate is determined after the data is randomly divided into training and test sets. In such a technique, it is most preferred to use 2/3/1/3 cross-validation, in which case 1/3 of the data is dropped and the remaining 2/3 is used to reconstruct the model .

  [0034] PLS or Partial Least Squares is a modeling algorithm that takes as input a matrix of predictors and a vector of monitoring scores and creates a set of prediction weights for each input predictor (Nguyen) Et al. (2002), Bioinformatics 18: 39-50). These prediction weights are then used to calculate a gene regulation score, which indicates whether each gene analyzed can predict a toxic response. As described in the Examples, this gene regulation score may then be used to calculate a toxicity reference prediction score.

[0035] From the nucleic acid hybridization data, a gene expression measure is calculated for one or more genes whose level of expression is detected in the nucleic acid hybridization value. As described above, this gene expression measure can include the RMA change rate. Toxicity reference score = Σw _i R ^FCi . “I” is the index number of each gene in the gene expression profile to be evaluated. “W _i ” is the PLS weight (or PLS score, see Table 2) for each gene. “R ^FCi ” is the RMA change rate for the i th gene determined from the normalized RMA matrix of gene expression data from the sample (above). Multiplying the PLS weight by the RMA change factor gives a gene regulation score for each gene, and summing the regulation scores for all its individual genes yields a toxicity reference prediction score for the sample or cohort of samples. To calculate a toxicity reference prediction score from at least one gene regulation score, or at least about 5, 10, 25, 50, 100, 500, or about 1,000 or more gene regulation scores Can do.

  [0036] In one embodiment of the invention, (a) at least one cell or tissue sample exposed to a toxin and nucleic acids for a plurality of genes of at least one cell or tissue sample exposed to the toxin vehicle Providing hybridization data; (b) converting hybridization data from at least one gene into a gene expression measure; and (c) a gene regulation score from the gene expression measure for the at least one gene. Prepare or create a toxicological model or toxicity model of the present invention by creating and (d) creating a toxicity reference prediction score for the toxin, thereby creating a toxicological model. This gene expression measure may be a gene change rate calculated by a logarithmic scale linear additive model, such as RMA, and a toxicity reference prediction score may be generated using PLS. This toxicity reference prediction score may then be added to a toxicity model or database, which may be used to predict at least one toxic effect of an unknown test agent or compound.

  [0037] In another preferred embodiment, this model is verified by techniques known to those skilled in the art. Preferably, cross-validation techniques are used. In such a technique, the data is randomly divided into training and test sets several times before determining the acceptable model success rate. In such a technique, it is most preferred to use 2/3/1/3 cross-validation, in which case 1/3 of the data is dropped and the remaining 2/3 is used to reconstruct the model .

Methods for predicting toxic effects [0038] Gene regulatory scores and toxicity prediction scores derived from cells or tissue samples exposed to toxins can be used to test or unknown drug or compound hepatotoxicity, nephrotoxicity, Alternatively, at least one toxic effect may be predicted including other tissue toxicities. Gene regulatory scores and toxicity prediction scores from cells or tissue samples exposed to the toxin may be used to predict whether a test agent or compound can induce a tissue pathology such as liver necrosis in the sample. . The toxicological prediction methods of the present invention are limited only by the availability of appropriate toxicological models and toxicological prediction scores. For example, by simply performing the new toxicological tests and models of the present invention using additional toxins, or agents that induce specific tissue pathologies, and appropriate cell or tissue samples, The prediction method of a given system such as a computer system or database can be expanded.

  [0039] As used herein, at least one toxic effect includes, but is not limited to, a detrimental change in the physiological state of a cell or organism. This reaction does not require it, but may involve certain pathologies such as tissue necrosis. Thus, this toxic effect includes effects at the molecular and cellular levels. For example, hepatotoxicity is an effect used herein, including but not limited to cholestasis, genotoxicity / carcinogenesis, hepatitis, human specific toxicity, hepatomegaly induction. , Steatosis, macroscopic steatosis, microvesicular steatosis, necrosis, non-genotoxic / non-carcinogenic toxicity, peroxisome proliferation, rat non-genotoxicity, and general hepatotoxic conditions.

  [0040] In general, an assay for predicting the toxicity of a test agent (or compound or multi-component composition) exposes a single cell or tissue sample, or a population of cells or tissue samples, to the test agent or compound. The test agent by assaying or measuring the level of relative or absolute gene expression of one or more of the genes, eg, one or more of the genes in Table 2 Providing nucleic acid hybridization data for at least one gene of one or more exposed cell or tissue samples, calculating a sample prediction score, and calculating the sample prediction score to one or more toxicities Comparing with a morphological reference score (see Example 1).

[0041] The sample prediction score may be calculated as follows. Sample prediction score = Σw _i R ^FCi . “I” is the index number of each gene in the gene expression profile to be evaluated. “W _i ” is the PLS weight (or PLS score) for each gene derived from a toxicity model. “R ^FCi ” is the RMA change rate for the i-th gene, determined from the normalized RMA matrix of gene expression data from the sample (above). Multiplying the PLS weights from a given model by the RMA change factor gives a gene regulation score for each gene, and summing the regulation scores for all its individual genes yields a predicted score for that sample.

  [0042] The nucleic acid hybridization data includes about (or at least) 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25 species. , 30, 50, 75, 100, 200, 500, 1000, or more genes, or a range of such numbers, eg, about 2-10, about 10-20, about 20 Of hybridization, including gene expression levels of sample nucleic acids with probes corresponding to -50, about 50-100, about 100-200, about 200-500, or about 500-1000 genes Any measurement can be included. Nucleic acid hybridization data for predicting toxicity includes measurements of almost all genes in a toxicity model. “Almost all” genes may be considered to mean at least 80% of genes in any one toxicity model.

  [0043] The methods of the invention for predicting at least one toxic effect of a test agent or compound may be performed by one individual, or at one location, or by multiple individuals, or multiple It may be implemented at the place. For example, the method of the present invention may be such that exposure of one test agent or compound to one or more cell or tissue samples is accomplished in one location, ie, nucleic acid processing, and the creation of nucleic acid hybridization data at another location. And a gene regulation score and a sample prediction score are calculated or generated elsewhere.

  [0044] In another embodiment of the invention, by administering a drug to a laboratory rat, a cell or tissue sample is exposed to a test drug or compound, and nucleic acids are processed from the selected tissue and hybridized to a microarray. To create nucleic acid hybridization data. This nucleic acid hybridization data is then sent to a remote server containing toxicological reference databases and software that enables the creation of individual gene regulation scores and one or more sample prediction scores from the nucleic acid hybridization data. . This software allows the user to pre-select specific toxicological models, and the generated sample prediction score to one or more toxicological references contained in such a score database. It may be possible to compare with a score. The user can then provide one or more appropriate outputs that provide or present results of data analysis, generation of gene regulation scores and sample prediction scores, and / or comparison with one or more toxicological reference scores. You may create or order a product.

  [0045] Data, including nucleic acid hybridization data, may be transmitted to the server by any available means, including secure direct dial-up or secure or unsecure internet connection. A toxicological prediction report, or any result of the methods herein, may be transmitted via that same mechanism. For example, an initial user may transmit nucleic acid hybridization data to a remote server via an internet link protected by a secure password, and then send a toxicological report from that server via that same internet link. You may request.

  [0046] The data transmitted by the remote user of the toxicity database or model may be raw, unnormalized data, or this data may be normalized from various background parameters prior to transmission. For example, data from the microarray may be normalized for various chips and background parameters such as those described above. This data can be in any form as long as the data can be recognized or properly formatted by the software in which it is available or provided as part of a database or computer system. For example, the microarray data may be provided and transmitted in a cel file or any other common data file created by analysis of microarray-based hybridization on commercially available technology platforms (eg available at www.affymetrix.com) (See Affymetrix GeneChip® Expression Analysis Technical Manual). Such files contain various information, such as, but not limited to, information related to customers or remote users, cell or tissue sample data or information, the hybridization technology or platform on which the data is created, and The test drug data or information may or may not be annotated.

  [0047] When receiving data, the nucleic acid hybridization data may be screened for database compatibility by any available means. In one embodiment, commonly available data quality management metrics can be applied. For example, using outlier analysis methods or techniques to identify samples that are incompatible with the database, e.g., samples that show false fluorescence values from control probes that are common between the data and the database or toxicity model. Good. In addition, various ones, including one or more disclosed in PCT / US03 / 24160, filed Aug. 1, 2003, claiming priority from US Provisional Patent Application No. 60/399727. Data QC metrics can be applied.

Cell or Tissue Sample Preparation [0048] As described above, a cell population exposed to a test agent, compound, or composition may be exposed in vitro or in vivo. For example, cultured or freshly isolated liver cells, particularly rat hepatocytes, may be exposed to the drug under standard laboratory and cell culture conditions. In another assay format, in vivo exposure may be achieved by administration of the drug to a living animal, such as a laboratory rat.

  [0049] Procedures for designing and conducting toxicity studies in vitro and in vivo are well known and are also known by Loomis et al., Loomis's Essentials of Toxology, 4th edition, Academic Press, New York, 1996; The Basics of Toxicity Testing, CRC Press, Boca Raton, 1992; Frazier, Ed., In Vitro Toxicity Testing, Marcel Dekker, New York. 1992 and many other texts on this subject.

  [0050] Two groups of test organisms are commonly used in in vitro toxicity tests. One group is a control and the other group receives the test compound at a single dose (for acute toxicity studies) or at a dosage regimen (for prolonged or chronic toxicity studies). In some cases, tissue extraction required by the methods of the present invention requires the test animals to be sacrificed, so if it is desirable to observe the dynamics of gene expression throughout the duration of the experiment, control groups and Both groups receiving the compound must be large enough so that the animals can be removed for tissue sampling.

  [0051] In designing toxicity studies, extensive guidance is provided in the literature on selecting appropriate test organisms for the compounds to be tested, routes of administration, dosage ranges, and the like. Water or saline (0.9% NaCl aqueous solution) is the solute selected for the test compound. This is because these solvents allow administration by various routes. When this is not possible due to solubility limitations, vegetable oils such as corn oil or organic solvents such as propylene glycol can be used.

[0052] Regardless of the route of administration, the volume required to administer a given dose is limited by the size of the animal used. It is desirable to keep the volume of each dose uniform within or between groups of animals. When using rats or mice, the volume generally administered by oral route should not exceed 0.005 mL per gram of animal. Even when aqueous solutions or saline solutions are used for parenteral injection, such solutions are usually considered harmless, but the volume allowed is limited. The intravenous LD ₅₀ of distilled water in mice is approximately 0.044 mL per gram of mouse and that of isotonic saline is 0.068 mL per gram of mouse. In some cases, the route of administration to the test animal should be the same or as similar as possible to the route of administration of the compound to humans for therapeutic purposes.

  [0053] When the compound is to be administered by inhalation, a special technique for generating test air is required. The method typically involves aerosolizing or nebulizing a liquid containing the compound. If the drug being tested is a liquid with appreciable vapor pressure, it can be administered by passing air through the solution under controlled temperature conditions. Under these conditions, the dose is estimated from the volume of air inhaled per unit time, the temperature of the solution and the vapor pressure of the drug involved. The gas is measured with a meter from the gas reservoir. When particles of solution are to be administered, the particles will not reach the terminal alveolar sac in the lung unless the particle size is less than about 2 μm. When administered by inhalation, a variety of devices and chambers are available to perform studies to detect the effects of stimulants or other toxic endpoints. The preferred method of administering a drug to an animal is through the oral route by intubation or by incorporating the drug in the feed.

  [0054] When an agent is exposed to a cell or cell culture in vitro, a cell population exposed to the agent can be divided into two or more identical aliquots, for example, by dividing the population into two or more identical aliquots. May be divided into subgroups. In some preferred embodiments of the methods of the invention, the cells exposed to the drug are derived from liver tissue. For example, cultured or freshly isolated rat hepatocytes are used.

  [0055] The methods of the present invention can generally be used to predict at least one toxic response and, as described in the Examples, the compound or test agent can be used to treat cholestatic hepatitis, genotoxicity / carcinogenesis. , Hepatitis, human specific toxicity, induction of liver hypertrophy, fatty liver, large fatty liver, small fatty liver, necrosis, nongenotoxic / noncarcinogenic toxicity, peroxisome proliferation, rat nongenotoxic toxicity, general It can be used to predict the likelihood of inducing various specific symptoms, such as liver toxicity, or symptoms associated with at least one other known toxin. The methods of the invention can also be used to determine the similarity of toxic responses to one or more individual compounds. In addition, the methods of the invention can be used to predict or elucidate cellular pathways that can be affected, induced or regulated by the compounds or test agents described above.

Databases and Computer Systems [0056] Databases and computer systems of the present invention are typically weighted index scores or PLS scores (see Table 2), gene regulatory scores, sample prediction scores, and / or toxicity of individual genes or toxicological markers. It includes one or more data structures that include a toxicity model or toxicological model as described herein, such as a model that includes a reference prediction score. Such databases and computer systems calculate or create the scores described herein, including the user manipulating the database contents or including individual gene regulation scores and sample prediction scores from nucleic acid hybridization data. It also includes software that makes it possible. To include gene or protein pathway information and / or to include information related to toxic mechanisms, including possible cellular and molecular mechanisms, the software allows users to determine toxicity, hepatotoxicity, nephrotoxicity It may be possible to predict, assay, or screen at least one toxic response, including and the like. As an example, the software may include Gene Logic ToxShield ™, such as software that includes at least one algorithm for converting hybridization data from various platforms, eg, from one microarray platform to another. At least one element from the predictive modeling system may be included. (See U.S. Provisional Application No. 60/613831, filed Sep. 29, 2004, which is incorporated herein by reference in its entirety for all purposes).

  [0057] As noted above, the database and computer system of the present invention provides an apparatus and device that allows access through a remote link, such as direct or direct dial-up access, or access via a password-protected Internet link. Software can be included.

  [0058] Any available hardware may be used to create the computer system of the present invention. Using any suitable computer platform, user interface, etc., necessary between sequence information, genes or toxicological marker information and any other information in the database or information provided as input A comparison may be performed. For example, many computer workstations are available from various manufacturers. Client / server environments, database servers, and networks are also available from various locations and are suitable platforms for the databases of the present invention.

  [0059] This database may be designed to include various parts, such as sequence databases and toxicological reference databases. Such databases and methods for the construction and construction of computer readable media containing such databases are available from a variety of locations, eg, 2002, which is incorporated herein by reference in its entirety. US Patent Application Publication No. 2003/0171876 (Patent Application No. 10/090144) filed on March 5, PCT Application Publication No. WO 02/095659 published on November 23, 2002, US Pat. No. 5,953727 See In one preferred embodiment, this database is Gene Logic, Inc. , Gaithersburg, MD, a ToxExpress® or BioExpress® database.

  [0060] Databases of the present invention are listed in GenBank (www.ncbi.nlm.nih.gov/entrez.index.html), KEGG (www.genome.ad.jp/kegg), PAR (www.grt.kyushhu-u Ac.jp/spad/index.html), HUGO (www.gene.ucl.ac.uk/hugo), Swiss-Prot (www.expasy.ch.sprot), Prosite (www.expasy.ch/tools/ may be linked to external or external databases such as snipstl.html), OMIM (www.ncbi.nlm.nih.gov/omim), and GDB (www.gdb.org). In a preferred embodiment, this external database is GenBank and its associated database maintained by NCBI (National Center for Biotechnology Information National Center for Biotechnology Information) (www.ncbi.nlm.nih.gov).

Toxicity or Toxicological Report [0061] As noted above, the methods, databases, and computer systems of the present invention can be used to generate, deliver, and / or transmit toxicity or toxicological reports. Consistent with the use of the terms “toxicity” and “toxicology” as used herein, “toxicity report” and “toxicology report” are interchangeable.

  [0062] Toxicity reports of the present invention typically contain information or data relating to the results of performing the inventive method. For example, as a result of performing a method for identifying at least one toxic effect of a test agent or compound described herein, the indication or prediction of at least one toxic response, such as toxicity, hepatotoxicity, nephrotoxicity, etc. A report describing the results of the method can be prepared or created. This report includes information related to toxic effects predicted by comparing at least one sample prediction score with at least one toxicity reference prediction score from the database, as well as a literature review or citation list, and / or one Other relevant information can be included such as information on the mechanism (s) of potential toxic effects. This report provides information entered by the database user and the method of the present invention, such as information about nucleic acid hybridization data, such as data integrity, and information used to annotate nucleic acid hybridization data. You can also.

  [0063] As an illustrative non-limiting example, the toxicity report of the present invention is PCT US 02, filed July 18, 2002, both of which are hereby incorporated by reference in their entirety for all purposes. And the report disclosed in US Provisional Application No. 60/613831 filed on Sep. 29, 2004, and US Patent Application No. 60/613831. As described elsewhere herein, this report may be generated by a server or computer system loaded with nucleic acid hybridization data by a user. A report related to the nucleic acid data may be generated and delivered to the user via remote means such as a password-secure environment available via the Internet or via available computer communication means such as email. .

Construction of Nucleic Acid Hybridization Data [0064] Any assay format that detects gene expression may be used to obtain nucleic acid hybridization data. For example, traditional Northern blotting, dot or slot blot, nuclease protection, primer-directed amplification, RT-PCR, semi-quantitative or quantitative PCR, branched DNA and differential display methods can Detection or nucleic acid hybridization data acquisition. Those methods are useful in some embodiments of the present invention. If fewer genes are detected, an assay based on high throughput amplification may be most efficient. However, the methods and assays of the present invention are most efficiently designed with hybridization-based methods that detect the expression of a large number of genes.

  [0065] Any hybridization assay format may be used to generate nucleic acid hybridization data, including solution-based and solid support-based assay formats. The solid support comprising oligonucleotide probes for the differentially expressed genes of the present invention can be a filter, a polyvinyl chloride dish, particles, beads, microparticles, or silicon or glass based chips, etc. . Such chips, wafers, and hybridization methods are available from various locations, such as those disclosed by Beattie (WO 95/11755).

  [0066] Any solid surface to which oligonucleotides can be attached, either directly or indirectly, covalently or non-covalently, can be used. Preferred solid supports are high density arrays or DNA chips. These contain specific oligonucleotide probes at predetermined positions on the array. The predetermined position may include more than one molecule of probe, but each molecule within that predetermined position has the same sequence. Such a predetermined position is called a feature. For example, there may be from 2, 10, 100, 1,000 to 10,000, 100,000, or more than 400,000 such features on a single solid support. The range in which the solid support or probe is attached is on the order of about 1 cm2. Probes corresponding to the genes in Tables 1-2 or from the above related applications can be attached to a single or multiple solid supports. For example, the probes can be attached to a single chip or to multiple chips that make up one chip set.

  [0067] Oligonucleotide probe arrays for expression monitoring, including bead assays or bead collection, can be made and used by any technique known in the art (see, eg, Lockhart et al., Nat. Biotechnol. (1996) 14,1675-1680; McGall et al., Proc.Nat.Acad.Sci.USA (1996) 93,13555-13460). Such probe arrays include at least two or more oligonucleotides that are complementary to or hybridize to the two or more genes listed in Table 2. For example, such an array has at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 50, 70, 100, 500 or 1000 or 1000 as described herein, or It may contain oligonucleotides that are complementary to or hybridize to further genes.

[0068] The sequences of expression marker genes in Table 2 are in public databases. Table 1 provides the SEQ ID NO and GenBank accession number (NCBI RefSeqID) for each of the sequences (see www.ncbi.nlm.nih.gov/) and also provides the cluster title for some of the genes is doing.
The sequences of genes in GenBank as of the filing date of the present application are expressly incorporated herein by reference in their entirety, and further related sequences, for example, sequences from the same gene of different length, The same applies to variant sequences, polymorphic sequences, genomic sequences and related sequences from different species (including human counterparts where appropriate).

  [0069] The term "background" or "background signal intensity" refers to non-specific between a labeled target nucleic acid and a component of an oligonucleotide array (eg, oligonucleotide probe, control probe, array substrate, etc.) Refers to a hybridization signal that results from free binding or other interactions. The background signal may also be caused by the internal fluorescence of the array component itself. A single background signal can be calculated for the entire array, or a different background signal can be calculated for each target nucleic acid. In a preferred embodiment, the background is calculated as the average hybridization signal intensity for the lowest 5% to 10% of the probes in the array, or if a different background signal is calculated for each target gene, each Calculated as the average hybridization signal intensity for the lowest 5% to 10% of the gene probes. Of course, those skilled in the art understand that if probes to a particular gene are well hybridized and appear to bind specifically to the target sequence, they should not be used for background signal calculations. Will do. Alternatively, a probe that is not complementary to any sequence found in the sample (eg, a probe directed to a nucleic acid of the opposite sense, or in a sample such as a bacterial gene when the sample is a mammalian nucleic acid) May be calculated as the intensity of the hybridization signal generated by hybridization to a probe directed to a gene not found in The background can also be calculated as the average signal intensity produced by the area of the array that lacks any probes.

  [0070] The expressions "specifically hybridize" or "specifically hybridize" are stringent when the sequences are present in the DNA or RNA of a complex mixture (eg, whole cells). Under conditions, refers to the binding, duplexing or hybridization of a molecule substantially or only to a specific nucleic acid sequence or sequences.

  [0071] As used herein, a "probe" binds to a complementary sequence target nucleic acid through one or more chemical bonds, usually through complementary base pairing, usually through hydrogen bond formation. Defined as a nucleic acid capable of As used herein, a probe comprises a natural base (ie, A, G, U, C or T) or a modified base (7-deazaguanosine, inosine, etc.). In addition, the base in the probe may be linked by bonds other than phosphodiester bonds, as long as it does not interfere with hybridization. Thus, the probe may be a peptide nucleic acid in which the constituent bases are connected by peptide bonds rather than phosphodiester bonds.

Nucleic Acid Sample [0072] A cell or tissue sample is exposed to a test agent in vitro or in vivo. When cultured cells or tissues are used, an appropriate mammalian cell extract, such as liver extract, may be added with the test agent to evaluate agents that may require biotransformation to show toxicity. . In a preferred manner, cultured cells of animal or human hepatocytes or primary isolates are used.

  [0073] The genes assayed by the present invention are typically in the form of mRNA or reverse transcribed mRNA. The gene may or may not be cloned. The gene may or may not be amplified. Cloning and / or amplification does not appear to bias the expression of genes within the population. However, in some assays it may be preferable to use polyA + RNA as the material since it can be used with fewer processing steps.

  [0074] As will be apparent to those skilled in the art, the nucleic acid samples used in the methods and assays of the invention may be prepared by any available method or process. Methods for isolating total mRNA are well known to those skilled in the art. For example, a method for isolating and purifying nucleic acids is described in “Laboratory Techniques in Biochemistry and Molecular Biology (Vol. 24) Hybridization with Nucleic Acid Probes: Theory and Nucleic Acid Probes, P. Tijsssen, Elsevier Press, N. Y. (1993) "is described in detail in Chapter 3. Such samples include RNA samples, but also include cDNA synthesized from mRNA samples isolated from cells or tissues of interest. Such samples also include DNA amplified from cDNA and RNA transcribed from the amplified DNA. One skilled in the art will recognize that it is desirable to inhibit or destroy the RNase present in the homogenate before it is used.

  [0075] A biological sample can be of any biological tissue, fluid, or cell from any organism and cells grown in vitro (eg, cell lines and tissue culture cells). Often, the sample is a tissue or cell sample that has been exposed to a compound, drug, medicament, pharmaceutical composition, potential environmental contaminant or other composition. In one manner, the sample will be a “clinical sample” that is a sample derived from a patient. Typical clinical samples include, but are not limited to, sputum, blood, blood cells (eg, white blood cells), tissue or fine needle biopsy samples, urine, peritoneal fluid or pleural effusion, or cells therefrom. . Biological samples may also include sections of tissue, such as frozen sections or formalin fixed sections taken for histological purposes.

Hybridization [0076] Nucleic acid hybridization involves contacting a probe with a target nucleic acid under conditions that allow the probe and its complementary target to form a stable hybrid duplex through complementary base pairing. Simply including. See WO99 / 32660. Nucleic acids that do not form hybrid duplexes are then washed away, allowing hybridized nucleic acids to be detected (typically through detection of attached detectable labels). It is generally accepted that nucleic acids are denatured by increasing the temperature of the buffer containing the nucleic acid or decreasing its salt concentration. Under low stringency conditions (eg, low temperature and / or high salt concentration), hybrid duplexes (eg, DNA: DNA, RNA: RNA or RNA: DNA) may be present even when the annealed sequences are not perfectly complementary. It is formed. Thus, the specificity of hybridization decreases with low stringency. Conversely, at higher stringency (eg, higher temperatures and / or lower salt concentrations), successful hybridization will tolerate fewer mismatches. One skilled in the art will recognize that hybridization conditions can be selected to provide any degree of stringency.

  [0077] In a preferred embodiment, hybridization is performed at low stringency (in this case 6x SSPET (0.005% Triton X-100) at 37 ° C) to ensure hybridization, and then Subsequent washing is performed at higher stringency (eg, 1 × SSPET at 37 ° C.) to remove mismatched hybrid duplexes. Subsequent washing may be performed with higher stringency (eg, lowering from 37 ° C. to 50 ° C. to 0.25 × SSPET) until the required level of hybridization specificity is obtained. Stringency can also be increased by the addition of agents such as formamide. By comparing hybridization to various controls that may be present (eg, expression level controls, normalization controls, mismatch controls, etc.) and hybridization to the test probe, hybridization specificity can be assessed.

  [0078] In general, there is a trade-off between hybridization specificity (stringency) and signal intensity. Thus, in a preferred embodiment, washing is performed at the highest stringency that yields consistent results and gives a signal intensity that exceeds background intensity. Thus, in a preferred embodiment, the hybridized array is continuously washed with higher stringency solution and read during each wash. Analysis of the resulting data set results in a wash stringency that no longer changes the hybridization pattern appreciably and provides the appropriate signal for the particular oligonucleotide probe of interest. Becomes clear.

Kit [0079] The present invention further provides a kit combining different combinations, high density oligonucleotide arrays, reagents used in the array, signal detection and array-processing equipment, toxicology database and analysis and the above database management software. Including. For example, the kit can be used to predict and model the toxic response of a test compound.

  [0080] Databases that may be packaged with this kit are described above. Specifically, the database software and information to package may include a database stored on a computer readable medium or transferred to a user's local server. In another format, the database and software information may be provided in a remote electronic format, such as a website, and its address may be packaged in a kit.

  [0081] Databases and software designed for use with microarrays are described in Balaban et al., US Pat. No. 6,229,911 (information collected from a few or many microarrays and stored as an indexed table. Computer-implemented method for managing data) and 6,185,561 (data that collects gene expression level data, adds additional characteristics, and reformats the data to answer various questions) Computer-based methods with mining capabilities). US Pat. No. 5,974,164 to Chee et al. Describes a software base for identifying mutations in nucleic acid sequences based on differences in probe fluorescence intensity between wild-type and mutant sequences that hybridize to a reference sequence. This method is disclosed.

  [0082] Without further explanation, one of ordinary skill in the art, using the foregoing description and the following illustrative examples, can make and utilize the compounds of the present invention and perform the claimed methods. Conceivable. Accordingly, the following examples specifically point out preferred embodiments of the present invention and should not be construed as limiting in any way the remainder of the disclosure.

Example 1: Construction of Toxicity Model Using RMA and LPS [0083] Various kidney toxins have been described in the art as described previously in the art and previously described in the above priority applications, protocols, And Sprague-Dawley male rats at various time points using and dosing methods. As illustrated by the protocol used, the toxin is administered, the animals are sacrificed, and kidney samples are taken at the following time points.

Observation of animals [0084] Clinical observation-Check mortality and mortality twice daily. Cage lateral observation (skin and hair, eyes and mucous membranes, respiratory system, circulatory system, autonomic and central nervous system, somatic movement pattern and behavior pattern). When signs of potential toxicity, including tremors, convulsions, salivation, diarrhea, somnolence, coma, or abnormal behavior or appearance, were recorded, including the time, extent, and duration of onset.

  [0085] 2. Physical examination-before randomization, before the first treatment and before sacrifice (slaughter).

  [0086] 3. Weight-before randomization, before first treatment and before sacrifice.

Clinical pathology [0087] Frequency-before autopsy.

  [0088] 2. Number of animals-one surviving animal.

[0089] 3. In between under anesthesia bleeding procedure _{-70% CO 2/30% O} 2, to obtain a blood pierced orbital sinus.

  [0090] 4. Collection of blood samples—About 0.5 mL of blood was collected in EDTA tubes for evaluation of hematology parameters. Approximately 1 mL of blood was collected in a serum separator tube for clinical chemistry analysis. Approximately 200 μL of plasma was obtained and frozen at −80 ° C. for test compound / metabolite estimation. An additional 2 mL of blood was collected in a 15 mL conical polypropylene vial and 3 mL of Trizol was immediately added. The contents were immediately vortexed and mixed upside down repeatedly. The tubes were frozen in liquid nitrogen and stored at -80 ° C.

Termination Procedure Termination Sacrifice [0091] At the above time points, the rats were weighed, examined physically, sacrificed by decapitation and blood loss. Animals were autopsied within approximately 5 minutes of sacrifice. The instrument was sterilized separately and a disposable one was used for each animal. Animal autopsy was performed according to a procedure approved by a board-certified pathologist.

  [0092] Animals that did not survive until termination were sacrificed without autopsy (following euthanasia due to carbon dioxide asphyxia if moribund). The approximate time of death of animals that were drowned or found dead was recorded.

Postmortem Procedure [0093] All tissues were harvested and frozen within approximately 5 minutes of animal death. Tissues were stored at approximately -80 ° C or stored in 10% neutral buffered formalin.

Tissue collection and processing [0094] Liver Inner right lobe-instantly frozen in liquid nitrogen and stored at -80 ° C.
2. Medial left lobe—stored in 10% neutral buffered formalin (NBF) and evaluated for gross and microscopic pathological examination.
3. Outer left lobe—frozen instantly in liquid nitrogen and stored at −80 ° C.

[0095] Heart 1. A sagittal cross-section containing two atrial parts and two ventricular parts was preserved in 10% NBF. The remaining heart was frozen in liquid nitrogen and stored at -80 ° C.

[0096] Kidney (both)
1. An incision was made in the left-half. Half was stored in 10% NBF and the remaining half was frozen in liquid nitrogen and stored at -80 ° C.
2. An incision was made in the right-half. Half was stored in 10% NBF and the remaining half was frozen in liquid nitrogen and stored at -80 ° C.

  [0097] Testes (both) —Sagittal cross sections of each testis were stored in 10% NBF. The remaining testis were frozen together in liquid nitrogen and stored at -80 ° C.

  [0098] Brain (all) —Cross section of cerebral hemisphere and diencephalon was stored in 10% NBF, and the rest of the brain was frozen in liquid nitrogen and stored at −80 ° C.

  [0099] The microarray sample was prepared with some modifications according to the protocol described in the Affymetrix GeneChip Expression Analysis Manual (Affymetrix GeneInc. Santa Clara, CA). Frozen tissue was ground into a powder using a Spex Certiprep 6800 Freezer Mill. Total RNA was extracted with Trizol (GibcoBRL) using the manufacturer's protocol. MRNA was isolated using an Oligotex mRNA Midi kit (Qiagen) and ethanol precipitated. Double-strand cDNA was generated from mRNA using the SuperScript Choice System (Invitrogen, Carlsbad CA). First strand cDNA synthesis was primed with T7- (dT24) oligonucleotide. The cDNA was extracted with phenol-chloroform and ethanol precipitated to a final concentration of 1 μg / mL. CRNA was synthesized from 2 μg of cDNA using Ambion's T7 MegaScript in vitro Transcribation Kit.

  [0100] Nucleotides Bio-11-CTP and Biol6-UTP (Enzo Diagnostics) were added to the reaction to label the cRNA with biotin. Following a 6 hour incubation at 37 ° C., impurities were removed from the labeled cRNA according to the Rneasy Mini kit protocol (Qiagen). cRNA was fragmented at 94 ° C. for 35 minutes (fragmentation buffer consisting of 200 mM Tris-acetic acid, pH 8.1, 500 mM potassium acetate, 150 mM magnesium acetate). According to the Affymetrix protocol, 55 μg of fragmented cRNA was hybridized with Affymetrix rat array set in a 45 ° C. hybridization oven at 60 rpm for 24 hours. The chips were washed and stained with streptavidin phycoerythrin (SAPE) (Molecular Probe) during the Affymetrix fluidics station. To amplify the staining, the SAPE solution was added twice with an anti-streptavidin biotinylated antibody (Vector Laboratories) staining step in between. Hybridization to the probe array was detected by fluorimetric scanning (Hewlett Packard Gene Array Scanner). Data were analyzed using Affymetrix GeneChip version 3.0 and Expression Data Mining (EDMT) software, GeneExpress database and S-Plus statistical analysis software (Insightful Corp).

Confirmation of Toxicity Markers and Toxicity Model Construction Using RMA and PLS Algorithm [00101] The RMA / PLS model is constructed as follows. A matrix of RMA change magnification expression values is created from DNA microarray data obtained from one or more tests. Such values can be derived, for example, from Irazary et al. (Nucl Acids Res31 (4): e15,2003) using the following formula: T (PM _ij ) = e _i + a _j + ε _ij Create according to the method. T represents a converted value that corrects the background, normalizes, and converts PM (perfect match) intensity to a logarithmic scale. e _i represents the log2 scale expression value found on the array i = 1−1, a _j represents the log scale affinity for probe j = 1−J, and ε _ij represents the error (various To compensate for differences in dispersion when using probes that bind with high intensity).

  [00102] In the RMA change rate matrix, rows represent individual fragments and columns are individual samples. A vehicle cohort median matrix is then calculated, where the rows represent fragments and the columns represent the vehicle cohort, one cohort for each test / timepoint combination. The values in this matrix are the median RMA expression values across the samples in such a cohort. Next, a matrix of normalized RMA expression values is created, where the rows represent individual fragments and the columns are individual samples. The normalized RMA value is the RMA value minus the value from the vehicle cohort median matrix corresponding to the time matched vehicle cohort. PLS modeling is then applied to the normalized RMA matrix (subset by obtaining a fragment as described below), with the surveillance score vector -1 = non-toxic, + 1 = toxic as the dependent variable, and Use the row of the normalized RMA matrix as the independent variable. PLS works by calculating a series of PLS components, where each component is a weighted linear combination of fragment values. A non-linear iterative partial least squares method is used to calculate the PLS component.

  [00103] To select the fragments, a vehicle cohort mean matrix is created, where the rows represent the fragments and the columns represent the vehicle cohort, with one cohort for each test / timepoint combination. is there. The values in this matrix are the mean RMA expression values across the samples in such a cohort. A treatment cohort mean matrix is then created, where the rows represent the fragments and the columns represent the treatment (non-vehicle) cohort, one cohort for each test / time point / compound / dose combination. . The values in this matrix are the mean RMA expression values across the samples in such a cohort. Next, a matrix of treatment cohort rates of change is created, where the rows represent fragments and the columns represent the treatment cohort, one cohort for each test / time point / compound / dose combination. The values in this matrix are the values in the processing cohort average matrix minus the values in the vehicle cohort average matrix that correspond to the appropriate vehicle cohort matched in time. A treatment cohort p-value matrix is then created, where the rows represent the fragments and the columns represent the treatment cohort, one cohort for each test / time point / compound / dose combination. The values in this matrix are p-values based on a two-sample t-test that compares the treatment cohort mean to the vehicle cohort mean corresponding to the appropriate time matched vehicle cohort. This matrix is converted to a binary code based on p-values less than 0.05 (code as 1) or greater than 0.05 (code as 0).

  [00104] Compute the sum of the rows of the binary processing cohort p-value matrix, where the sum of the rows represents the "gene regulation score" for each fragment, compared to the vehicle cohort that is time matched Presents the total number of treatment cohorts in which the fragment showed differential regulation (up or down regulation). Then, PLS modeling and 2/3/1/3 cross-validation, obtaining the top N fragments according to their regulatory scores, changing the number of N and PLS components, and recording the model success rate for each combination Implement based on what to do. N is selected to be a point that minimizes the error rate due to cross-validation. In the PLS model, each of these N fragments receives a PLS weight (PLS score) that corresponds to the usefulness or predictive ability of the fragment in this model (see Table for an exemplary list of PLS scores for the kidney general toxicity model). 2).

Example 2: Method for predicting at least one toxic effect of a test agent [00105] Determining whether a sample from an animal treated with a test agent or compound exhibits at least one toxic or toxic response For this purpose, RNA is prepared from a cell or tissue sample exposed to the drug and hybridized with a DNA microarray as described in Example 1 above. From the nucleic acid hybridization data, a predicted score for the sample is calculated according to the following formula and compared to a reference score from a toxicity reference database. Sample prediction score = Σw _i R ^FCi . “I” is the index number of each gene in the gene expression profile to be evaluated. “W _i ” is the PLS weight for each gene (or PLS score, see Table 2 for an exemplary list of PLS scores for the kidney general toxicity model). “R ^FCi ” is the RMA change rate for the i-th gene, determined from the normalized RMA matrix of gene expression data from the sample (above). Multiplying the PLS weights by the RMA change rate gives a gene regulation score for each gene, and summing the regulation scores for all its individual genes gives a predicted score for that sample.

  [00106] As a quality control (QC) check, an average correlation evaluation is performed for each input test. After creating the RMA matrix (gene for each sample), the Pearson correlation matrix between samples is calculated. This matrix is for each sample. For each sample matrix row, calculate the average of all correlation values in that row of the matrix, excluding the diagonal term (always 1). This average is the average correlation for that sample. If the average correlation is below a certain threshold (eg, 0.90), the sample is flagged as a potential outlier. This process is repeated for each row (sample) of inspection. Outliers flagged by the mean correlation QC check are removed from any normalization, prediction, or compound-like steps downstream in this process.

  [00107] To establish a cutoff value for the toxicity prediction score for a toxicity model, using scores from all toxic and non-toxic samples in the training set, the true positive rate for the cutoff value for each possible score and Calculate the false positive rate. This creates an ROC curve that is used to set a cutoff score at a point in the ROC curve that corresponds to a false positive rate of about 5%. For example, in the nephrotoxicity model of Table 2, the cut-off prediction score is about 0.318. If the sample score is about 0.318 or greater, after exposure to the test compound, the sample can be predicted to show a toxic response. If the sample score is below 0.318, it can be predicted that the sample will not show a toxic response.

  [00108] By setting a score of -1 for each gene for which a toxic response cannot be predicted and by setting a score of +1 for each gene for which a toxic response can be predicted The model can be trained. Cross validation of the RMA / PLS model may be performed by the compound-drop method and by the 2/3: 1/3 method. In the compound drop method, sample data from animals treated with a particular test compound is taken from a model and compared to a model containing the entire data set to determine if this model can predict toxicity. The 2/3: 1/3 method takes gene expression information from a random one-third gene in the model and compares whether this subset model can predict toxicity with a model that includes the entire data set .

  [00109] Compound similarity is evaluated as follows. In the same manner as described above, the cohort change rate vector for each test / time point / compound / dose combination is calculated. This vector is only the fragment used in the PLS prediction model. Next, using each cohort vector in the reference database, a Pearson correlation for the cohort change magnification vector is calculated (this is also only a fragment used in the PLS prediction model). Finally, classify these Pearson correlations from highest to lowest and report the results.

  [00110] A report may be generated that includes information or data related to the results of a method for predicting at least one toxic effect. The report can include information related to toxic effects predicted by comparison of at least one sample prediction score with at least one toxicity reference prediction score from a database. This report provides information entered by the database user and the method of the present invention, such as information about nucleic acid hybridization data, such as data integrity, and information used to annotate nucleic acid hybridization data. You can also. See PCT US 02/22701 for a non-limiting example of a toxicity report that may be generated.

Example 3: Converting RMA data from one platform to another [00111] Developing an algorithm to convert probe intensity data from a first type of microarray to RMA data of a second type of microarray did. This is beneficial to the customer as it gives the customer the freedom to select the type of microarray that is desired to be used in the RMA / PLS prediction model. This is often the latest microarray on the market. This algorithm is beneficial for populations building RMA / PLS statistical models on microarray data, as there is no need to expand funds and resources to reconstruct statistical models built on discontinued microarrays. is there.

  [00112] Developed transformation algorithms can be used on data, from Affymetrix GeneChip® rat RAE2.0 microarrays to Affymetrix GeneChip® rat RGU34 A microarray data. This transformation also allows for the prediction of customer data generated on the RAE 2.0 microarray platform using the RMA / PLS toxicity genomics model built on the Affymetrix RGU34 A microarray platform. This conversion algorithm was tested using the liver toxicity model described in US Provisional Patent Application No. 60/559949 and incorporated herein by reference.

  [00113] The first step to the use of the conversion algorithm is to map the microarray fragments. The RGU34 A microarray fragment containing the liver toxicity model was mapped to the RAE2.0 microarray. The liver toxicity model is based on 1,100 Affymetrix GeneChip® RGU34 A microarray fragments. Of the 1,100 fragments in this model, 907 were suggested by Affymetrix to match the fragments in the RAE2.0 microarray. See Affymetrix's “User's Guide to Product Comparison Spreadsheets”, incorporated herein by reference. Another 105 fragments were located in the fragment sharing the same RefSeq ID, and 55 fragments were located in the fragments located in the same Unigene cluster. 1067 positioning fragments were reduced to 1053. The 1053 positioned fragments presented 16 RGU34 A and 11 RAE2.0 probes. The 47 fragments that were not located in the RAE2.0 microarray were assigned a RMA change factor of 0 for all samples and this fragment did not contribute to the prediction.

[00114] After locating the microarray fragment, a training sample is selected and a transform model weight is calculated. We searched Gene Logic's ToxExpress® reference database, a database built on the Affymetrix RGU34 A platform, for samples over a large quartile range for signal intensity. Since we have previously determined that this sample selection method is reliable in the development of human sample conversion algorithms, we selected samples over a large variable space. The sample maximized Σ _i (Max (X _ij ) −Min (X _ij )). In the formula, i indicates a gene and j indicates a sample.

  [00115] The inventors have found that the sample size calculation is stable with sampling of approximately 100 microarrays. For this reason, a training set consisting of 100 compounds from rat liver tissue and vehicle was selected.

  [00116] 100 training samples were used to train weights in the transformation algorithm. This step is important because it provides a quantitative aspect of the conversion. Weight training was performed based on multiple regression analysis using probe values as independent variables and RMA expression as the sum of dependent variables.

[00117] Test samples were evaluated using a trained transformation algorithm. A multiple regression model was constructed based on 11 perfect match probe intensities and a predicted RGU34 expression value was generated from the weighted sum of RAE2.0 probe values. Each test assay was adjusted to an average probe intensity of 10 (log scale). The conversion algorithm used is obtained from the following equation.
Y _i ^RGU34 = β _io + Σβ _ij LOG (X _ij ^RAE2.0 / S)
Where Y is the RGU34 RMA expression value for the fragment, i = 1. . . 1053, j = 1. . . X _ij ^RAE2.0 in case of 11 is the perfect match probe intensity value for the marker gene in the RAE2.0 microarray, and S is the chip scale factor Σ _ij X _ij ^RAE2.0 / n. The probe intensity was lowered to a minimum intensity value of 30.

  [00118] An alternative approach to using multiple regression models exists to convert RAE2.0 data to RGU34 RMA data. Non-linear regression on probe values as well as canonical correlation of RAE2.0 probe with RGU34 A probe can be used. RMA values in the RAE2.0 microarray can be calculated and then normalized to the RGU34 A RMA value or normalized to its displacement value. Furthermore, the multiple regression analysis used in this example does not take into account mismatched probes, but an analysis that takes into account mismatched probes can be used.

  [00119] The liver prediction model was used to compare the prediction results of the test data from the RGU34 microarray with the test data derived from the transformed RAE2.0 array data. The consistency between the RGU34 array results and the transformed RAE2.0 array results is very high. Table 3 provides the number of test samples per compound predicted to be toxic among the total number of samples for compounds using RGU34 RMA data and RAE2.0 converted RMA data. Amitriptyline, estradiol, amiodarone, diflunisal, phenobarbital, dioxin, ethionine, and LPS were selected as test toxicants. Clofibrate was chosen because it is a rat-specific toxicant. Metformin, rosiglitazone, chlorpheniramine, and streptomycin were selected as test negative controls. This rat-specific toxicant and all tested negative controls accurately predicted non-toxicity.

Example 4: Database [00120] A web-based software predictive modeling system called ToxShield ™ Suite consisting of a collection of RMA / PLS toxicity predictive models was created. A liver RMA / PLS prediction model was constructed to allow the user to identify and classify various toxic and mechanical responses to unknown or test compounds. This model includes general toxicity, necrosis, steatosis, macrodropleosis, microvesicular steatosis, cholestasis, hepatitis, carcinogenicity, carcinogenicity due to genotoxicity, carcinogenicity due to nongenotoxicity, rat-specific non-inheritance A wide variety of indicative pathologies and signs are presented, including carcinogenicity due to toxicity, peroxisome proliferation, and inducer / hepatomegaly. The results of the toxicity model can present a detailed classification of the test compound or unknown compound and infer its mechanical information. Current models available as part of this software system are related to liver toxicity, including but not limited to liver primary cell culture, kidney, heart, spleen, bone marrow, and brain Models for specific toxicity of other organs can be used.

  [00121] The conversion algorithm described in Example 3 can be implemented in a software product such as ToxShield ™ Suite. The customer enters that data created on a microarray, such as an Affymetrix RAE2.0 GeneChip® microarray platform. The software uses an algorithm to convert customer gene expression data into RMA data that is compatible with software toxicity genomics models built exclusively on a second microarray platform such as the Affymetrix RGU34 A GeneChip® microarray. Convert. The prediction model can then be used to create visualizations and predictions from the customer data.

  [00122] Although the invention has been described in detail in connection with the above examples, it should be understood that various modifications can be made without departing from the spirit of the invention. Accordingly, the invention is limited only by the claims. All cited patents, patent applications, and publications mentioned in this specification are hereby incorporated by reference in their entirety.

Claims (55)

A method for predicting at least one toxic effect of a test agent comprising:
(A) providing nucleic acid hybridization data for a plurality of genes in at least one cell or tissue sample exposed to the test agent;
(B) converting the hybridization data from at least one gene into a gene expression measure;
(C) generating a gene regulation score from the gene expression measure for the at least one gene;
(D) creating a sample prediction score for the drug;
(E) comparing the sample prediction score to a toxicity reference prediction score, thereby predicting at least one toxic effect of the test agent.   The method of claim 1, wherein at least one cell or tissue sample is exposed to a test drug vehicle.   3. The method of claim 2, wherein the conversion of step (b) comprises normalizing the hybridization data for background hybridization and expression induced by the test agent vehicle.   The method of claim 2, wherein the gene expression measure is a gene change rate.   5. The method of claim 4, wherein the rate of change is calculated by a logarithmic scale linear additive model.   6. The method of claim 5, wherein the logarithmic scale linear additive model is a robust multi-array average (RMA).   The method of claim 1, wherein the nucleic acid hybridization data is screened by a quality control process that measures outlier data.   The method of claim 1, wherein step (c) includes dimensionality reduction using partial least squares (PLS).   The method of claim 1, wherein the sample prediction score is created using a weighted index score for each gene.   The method of claim 1, wherein a sample prediction score for the drug is generated from a gene regulation score for the at least one gene.   11. The method of claim 10, wherein a sample prediction score for the drug is generated from a gene regulation score for at least about 10 genes.   11. The method of claim 10, wherein a sample prediction score for the drug is generated from a gene regulation score for at least about 50 genes.   11. The method of claim 10, wherein a sample prediction score for the drug is generated from a gene regulation score for at least about 100 genes. The toxicity reference prediction score,
(A) providing nucleic acid hybridization data for at least one cell or tissue sample exposed to a toxin and a plurality of genes of the at least one cell or tissue sample exposed to the toxin vehicle;
(B) converting the hybridization data from at least one gene into a magnification of change;
(C) creating a gene regulation score from the fold change factor for the at least one gene;
And (d) creating a toxicity reference prediction score for the toxin.   The method of claim 1, wherein step (a) comprises loading nucleic acid hybridization data to a server via a remote connection.   The method of claim 15, wherein the remote connection is through the Internet.   The method of claim 1, wherein the toxicity reference prediction score is included in a database.   The method of claim 17, wherein the toxicity reference prediction score is derived from a toxicological model.   19. The method of claim 18, wherein the toxicological model is selected from the group consisting of an individual toxin model, a toxin class model, a general toxicological model, and a tissue pathology model.   The method of claim 1, further comprising: (f) generating a report that includes information related to the toxic effect.   21. The method of claim 20, wherein the report includes information related to the mechanism of toxic effects.   21. The method of claim 20, wherein the report includes information related to a toxin used to prepare the toxicity reference prediction score.   21. The method of claim 20, wherein the report includes information relating to at least one similarity between the test agent and a toxin.   The method of claim 16, wherein the hybridization data is included in a plain text file.   The method of claim 16, wherein the hybridization data is included in a CEL file.   The method of claim 1, wherein the nucleic acid hybridization data is annotated with information selected from the group consisting of customer data, cell or tissue sample data, hybridization technology data, and test drug data.   16. The method of claim 15, wherein step (a) further comprises selecting at least one toxicity model for predicting the at least one toxic effect. A method for providing a report including a prediction of at least one toxic effect of a test agent comprising:
(A) Nucleic acid hybridization data relating to a plurality of genes of at least one cell or tissue sample exposed to the test agent and at least one cell or tissue sample exposed to the test agent vehicle via a remote link Receiving to the server
(B) converting the hybridization data from at least one gene into an RMA (Robust Multi-Array Average) rate of change;
(C) generating a gene regulatory score from the RMA change rate for the at least one gene;
(D) creating a sample prediction score for the drug;
(E) comparing the sample prediction score to a toxicity reference prediction score;
(F) providing a report including information relating to the at least one toxic effect. A method for creating a toxicological model comprising:
(A) providing nucleic acid hybridization data for a plurality of genes in at least one cell or tissue sample exposed to a toxin;
(B) converting the hybridization data from at least one gene into a gene expression measure;
(C) generating a gene regulation score from a gene expression measure for the at least one gene;
(D) creating a toxicity reference prediction score for the toxin and thereby creating a toxicological model.   30. The method of claim 29, wherein at least one cell or tissue sample is exposed to a test drug vehicle.   30. The method of claim 29, wherein the conversion of step (b) comprises normalizing hybridization data for background hybridization and expression induced by the test agent vehicle.   30. The method of claim 29, wherein the gene expression measure is a gene change rate.   33. The method of claim 32, wherein the rate of change is calculated by a logarithmic scale linear additive model.   34. The method of claim 33, wherein the logarithmic scale linear additive model is a robust multi-array average (RMA).   30. The method of claim 29, wherein creating step (c) includes dimensionality reduction using partial least squares (PLS).   30. The method of claim 29, wherein step (d) comprises creating a weighted index score for each gene.   30. The method of claim 29, wherein a toxicity reference prediction score for the toxin is generated from a gene regulation score for the at least one gene.   38. The method of claim 37, wherein a toxicity reference prediction score for the drug is generated from a gene regulation score for at least about 10 genes.   38. The method of claim 37, wherein a toxicity reference prediction score for the drug is generated from a gene regulation score for at least about 50 genes.   38. The method of claim 37, wherein a toxicity reference prediction score for the drug is generated from a gene regulation score for at least about 100 genes.   30. The method of claim 29, wherein the toxicological model is selected from the group consisting of an individual toxin model, a toxin class model, a general toxicological model, and a tissue pathology model.   30. The method of claim 29, further comprising validating the model.   43. The method of claim 42, wherein the confirmation comprises using a cross-validation procedure.   44. The method of claim 43, wherein the cross-validation procedure is a 2/3/1/3 confirmation procedure. (A) a computer readable medium comprising a toxicity model for predicting the toxicity of a test agent produced by the method of claim 29;
(B) a computer system including software that allows a user to predict at least one toxic effect of the test agent by comparing a sample prediction score to a toxicity reference prediction score in the toxicity model.   The software allows a user to compare quantitative gene expression information obtained from a cell or tissue sample exposed to a test agent with quantitative gene expression information in the toxicity model, so that the test agent is a toxin. 46. The computer system of claim 45, which makes it possible to predict whether there is any.   Further comprising software that allows a user to transmit nucleic acid hybridization data from a cell or tissue sample exposed to a test agent from a remote location to predict whether the test agent is a toxin. Item 46. The computer system according to Item 45.   The computer system according to claim 45, wherein nucleic acid hybridization data from the sample can be transmitted via the Internet.   46. The computer system of claim 45, wherein the nucleic acid hybridization data is microarray hybridization data.   46. The computer system of claim 45, wherein the nucleic acid hybridization data is PCR data.   46. The computer system of claim 45, further comprising a data structure comprising at least one toxicity reference prediction score.   46. The computer system of claim 45, wherein the data structure further comprises at least one gene PLS score.   46. The computer system of claim 45, wherein the data structure further comprises at least one gene regulation score.   46. The computer system of claim 45, wherein the data structure further comprises at least one sample prediction score.   A computer readable medium comprising a data structure including at least one toxicity reference prediction score and software for accessing the data structure.