SE517259C2 - Molecular identification system - Google Patents

Abstract

Mass data are typically not unique -I.e., each experimentally determined mass can match randomly one or several molecules in a database. Random matching between mass data and molecules in a database can cause false identification results. In order to minimize false results, random matching must be appropriately accounted for in a method for molecule identification. The invention provides a method to determine, for any molecule in a database and for any experimental and database search constraints, the probability that a particular number of matches between the mass data and masses of molecule constituents results from random matching. The method utilizes the determined probability for random matching to assign scores and rank molecules in a database. The invention further provides a method of generating a frequency function of scores for any experimental condition or database search constraints, wherein the scores relate to random identifications of molecules. Frequency functions are necessary and sufficient tools for testing the significance of a score associated with an identification of an unknown biological molecule.

Description

25 30 5 1 '2 5 9 2 proteins in a database. The protein is identified based on an evaluation of one or both of these comparisons.

Mass spectrometry determines a peptide mass m, - to an inaccuracy of: |: Am, -, with Ami / m, - typically> 30 ppm. Within the mass range m, d: Am, - proteolytic peptide masses of fl your proteins in a genome database can match. Thus, an unmodified peptide will randomly match your proteins in the database, in addition to the true match with the protein actually found in the sample, and, a modified peptide will give only random matches. As a consequence, a database search using mass spectrometric information will not always unambiguously identify a protein. Therefore, in order to perform accurate and reliable molecular identification, instruments for obtaining mass data must be appropriately linked with the use of other technical resources for comparing mass data and mass data obtained from a database. The link can be a system that uses a method that includes ways of comparing data and database information, conveniently run via a computer.

Despite the rapidly increasing pressure of influence of mass spectrometric protein identification on proteome research, the problem of carefully taking into account the phenomenon of random mass conformity in a database search system has been overlooked. When processes with increased complexity are explored with MS-based protein identification, the use of optimized procedures will become critical.

An optimized protein identification system cannot be designed without due regard to the process of random mass matching.

Background Art Identification of proteins by the approach described above requires a scheme for determining the best match between experimental data and a sequence in the database. Existing schemes for determining the best match include ranking by the number of matches (W. J. Henzel et al., Proc. Natl. Acad. Aci. USA 90, 5011, 1993), a scoring system based on the observed frequency of peptides from all proteins in a database within a given molecular weight range (the so-called “MOWSE score”) (DCJ Pappin et al., 10. 15 20 25 30 517 259 s Current Biology 6, 327, 1993)), and a scheme based on Bayesian probabilities (W. Zhang et al., Anal. Chem. 72, 2482, 2000).

None of these schemes take into account the problem of random conformity in an appropriate way. The lack of appropriate consideration for random conformity prevents optimal performance in molecular identification procedures, as random conformity can cause false identification results - especially when the quality of mass spectrometric data is poor.

SUMMARY OF THE INVENTION The object of this invention is to overcome the shortcomings of the above schemes, i.e. to provide a method which solves the problem of random conformity.

This and other intentions have been achieved by providing a system including methods for determining the probability that a particular score is due to random matching and using the calculated probability to rank molecules. The method comprises a) determining the number of matches between a database molecule and mass data; b) calculating the probability that a database molecule would yield a certain number of matches by chance; c) calculation of a score based on one or fl your probabilities calculated in b; d) comparing points for molecules in the database and e) identifying the molecule or molecules that give the best point or scores.

The invention further provides a method for generating a frequency function for the number of matches for random (false) molecular identifications under any experimental condition. The method comprises: a) defining a subpopulation of molecules included in the database; b) calculating the probability that a molecule in this subpopulation would give a certain number of coincidences by chance; c) calculating the probability that all molecules in the subpopulation would give a maximum of a certain number of coincidences by chance; d) calculating the probability that at least one molecule in the subpopulation would give at least a certain number of coincidences by chance; and e) determining the relative frequency of each number of matches using the probability calculated in step (d) for each number of matches, to generate therefrom a frequency function for the number of random matches at random protein identification.

Brief Description of the Drawings Fig. 1 shows frequencies (ie, number of corresponding proteins) for different tryptic masses of pulps in a database.

Fig. 2 shows mass distribution peaks for tryptic peptides.

Fig. 3 shows the performance of an implementation of an embodiment of the invention compared to prior art protein identification systems. The diagram shows results from simulations using the invention (called Probity), a Bayesian method and a method based on the number of matches. i Detailed description of the invention Many applications of molecular identification are by nature large-scale. Examples of large-scale molecular identification are found in protein projects, where thousands of proteins from cells are to be identified, or cells are scanned for molecular markers of disease states. The ultimate goal of molecular identification procedures is to rely on simple, fast and automated procedures and instruments. The technical solutions of the system that links and compares data with database information are extremely important for the design of automatic molecular identification instruments, as the system used will strongly influence the ability to obtain a high frequency of true identification results, which is particularly critical when data quality is poor. Furthermore, automatic identification instrumentation requires that the quality of identification results can be assessed automatically through the use of a signal test (J. Eriksson m fl, Anal. Chem. 72, 999, 2000). However, a reliable automatic protein identification system cannot be designed without regard to, or with inappropriate regard to, the process of random mass conformity.

One object of this invention is to provide a system that utilizes procedures that allow for more accurate molecular identification and more accurate and rapid signing testing of identification results. The method according to the invention takes due account of the phenomenon of random conformity, and is therefore well suited for implementation in an automatic molecular identification system.

A special concern regarding large-scale molecular identification is the time required to obtain an identification result together with a quality determination of the result. A quality determination can be completed with a signi fi test, which requires knowledge of functions that describe points for false results. Such frequency functions are currently obtained by simulating random molecular identification. However, since the time required to derive a frequency function by means of simulation is approximately 1000 times longer than when using this invention, there is a need to derive such a frequency function from an analytical formula. In one embodiment of the invention there is provided the provision of an analytical formula for the derivation of a frequency function. The methods of this invention are well suited for, but not limited to, the application in which the molecules are biological molecules that exist in the cells of an organism.

Biological molecules include any biological polymer that can be degraded into constituents. The decomposition into constituents should preferably be at predictable positions to form predictable masses. Examples of biological molecules include proteins, nucleic acid molecules, polysaccharides and carbohydrates.

An experimental biological molecule is a biological molecule that must be identified; the experimental biological molecule can also be called an unknown biological molecule. A theoretical biological molecule is a biological molecule is a known biological molecule described in a database.

Proteins are polymers of amino acids. Components of proteins include amino acids. A protein contains approximately at least ten amino acids, but more preferably at least 50 amino acids and even more preferably at least 100 amino acids.

Nucleic acids are polymers of nucleotides. Nucleic acid constituents include nucleotides. Typically, a nucleic acid contains at least 100 nucleotides, but more preferably at least 500 nucleotides. Polosaccharides are polymers of monosaccharides. Components of polysaccharides include one or two monosaccharides. Typically, a polysaccharide contains at least five monosaccharides, but more preferably at least ten monosaccharides.

Mass data from biological molecules are quantifiable information about the masses of the constituents of the biological molecule. Mass data includes individual mass spectra and groups of mass spectra. Mass spectra can be in the form of peptide maps, oligonucleotide maps or oligosaccharide maps.

The method of this invention involves generating experimental mass data for the experimental molecule within a certain mass range. Mass data includes the measured masses. The method also includes generating theoretical mass data in the same mass range. In a concrete form, mass data is a subset of the experimental mass data.

For example, mass data for molecules can be generated in any way that provides mass data within a certain inaccuracy. Examples include matrix-assisted laser desorption / ionization mass spectrometry, electrospray ionization mass spectrometry, chromatography and electrophoresis. Mass data can also be generated by a computer configured with software or otherwise.

For the purposes of this invention, mass data is determined, e.g. a peptide mass, to some inaccuracy: hAmb rather so Ami / m, - <10,000 ppm, preferably <100 ppm, and most preferably <30 ppm.

One step in the generation of mass data for a molecule may involve first cleaving the molecule into components. Biological molecules can be cleaved by methods known in the art. Biological molecules are preferably broken down into constituents in predictable positions to form predictable masses. Cleavage procedures include chemical methods. Biological molecules can be degraded by contacting the biological molecule with any chemical substance.

For example, proteins can be predictably degraded to peptides by cyanogen bromide or enzymes such as trypsin, endoproteinase Asp-N, V8 protease, endoproteinase Arg-C, etc. Nucleic acids can be degraded in a predictable manner to 10 constituents. with restriction endonucleases such as Eco RI, Sma I, BamH I, Hinc II, etc. Polysaccharides can be degraded into constituents by enzymes such as maltase, amylase, alpha-mannosidase, etc.

In this invention, a mass range (mmm, mmax) is determined for the experimental data. The mass range can be the mass range for the experimental data. In one embodiment, the mass range is the minimum and maximum of measured masses of experimental mass data for a molecule.

A molecule database is any collection of information about the characteristics of hay molecules. A molecule database can be a database of biological molecules.

Databases are the preferred method for storing both amino acid sequences of polypeptides and of the nucleic acid sequences encoding its polypeptides. Different types of databases exist and these have advantages and disadvantages when considered with respect to the hypothesis of a polypeptide identification experiment.

While the database entry for an amino acid sequence can be designed as a simple text fi l for a user to look up a particular polypeptide, many p databases are organized so that they constitute ib excel complicated structures. The detailed implementation of the database in a particular system may be based on a collection of simple texts (database), a collection of tables (a relational database), or it may be organized around concepts derived from the perception of a protein, a gene or an organism (an object-oriented database).

Protein mass data can be predicted from nucleic acid sequence databases. Alternatively, protein mass data can be predicted directly from protein sequence databases containing a collection of amino acid sequences represented by a strand of one-letter or three-letter codes for the residues in a polypeptide, beginning at the N-terminus of the sequence. These codes may contain non-standard characters to indicate ambiguity in a particular position (such as "B" indicating that it may be "D" (aspartic acid) or "N" (asparagine)). The sequences typically have unique number-letter combinations associated with them, which are used internally in the database to identify the sequence, which is commonly referred to as the "accession number" for the sequence. 10 15 20 25 30 517 2 se - 8 Databases may contain a combination of amino acid sequences, comments, literature references and notes regarding known post-translational modifications of the sequence. A database that contains these elements is called annotated. Annotated databases are used if any functional or structural information is known for the mature protein, as opposed to a sequence known only from the translation of a piece of nucleic acid sequence. Non-annotated databases contain only the sequence, an access number and a descriptive title.

The known background information of an experimental molecule by which the database search can be delimited can include any information.

Some examples of background information are information on the category of an experimental biological molecule, knowledge of or an assumption about the mass of the experimental biological molecule and the isoelectric point of the experimental biological molecule.

For example, the observed molecular mass or the observed isoelectric point of a protein can be used in combination with the measured peptide masses generated by proteolysis, to delimit the search for a polypeptide. In particular, the comparison between the theoretical mass data for the database proteins and the mass data for the unknown protein can be limited to only those proteins in the database that are within a selected mass range. The selected mass range is preferably within 50% of the mass of the unknown protein, but even more preferably within 35% and most preferably within 25%. Similarly, the comparison between the theoretical mass data of the database proteins and the mass data of the unknown protein can be limited to only those proteins in the database that are within a certain selected range of the isoelectric point. The isoelectric point (pI) of a protein is the pH at which its net charge is zero. The selected range of the isoelectric point is preferably within 50% of the isoelectric point of the unknown protein, but more preferably within 35% and most preferably within 25%.

In addition, further information, such as a protein sequence, about an experimental biological molecule can be obtained by generating fragment mass data for the experimental and theoretical biological molecules. Fragment mass data for a peptide can be generated in any way that provides fragment mass data within a certain inaccuracy. Experimental conditions include the form of energy used to generate fragment mass data. Vibration excitation can be used. The electronic excitation can be generated by collisions between the peptide and electrons, photons, gas molecules (eg argon) or a surface.

In another example, the experimental fragment mass spectrum of a peptide from an enzymatically cleaved unknown protein is compared with the theoretical masses calculated by applying the rules of the enzyme specificity and the rules of fragmentation known to those skilled in the art to the amino acid sequence of a database protein.

Fragment mass data for the purposes of this invention can be generated using multidimensional mass spectrometry (MS / MS), also known as tandem mass spectrometry. quadrupole ion fal- _ lemass spectrometer. A single peptide from a cleaved protein is subjected to an MS / MS experiment and the observed pattern of fragment ions is compared with the patterns of fragment ions predicted from database sequences.

In one embodiment, the invention provides a method for determining the probabilities that the score that a particular molecule in a database can give by chance when compared to mass data. The procedure can work under various experimental and database search restrictions. The score can be the number of matches between masses derived from known or hypothetical molecules or molecular constituents in a database and mass data from one or more known or unknown molecules or molecular constituents. The score can also result from a calculation that uses the number of matches.

In one embodiment, the invention provides a method of extracting information about molecules in a database. Examples of information that can be extracted from a database are total molecular mass, charge, isoelectric point, hydrophobicity and known or hypothetical chemical modifications, as well as mass, charge, isoelectric point, hydrophobicity and known or hypothetical chemical modifications of molecular constituents. In this embodiment, the invention provides a method of performing actions with molecules in the database that are intended to mimic events that occur in an experiment. Examples of events are the decomposition of molecules into molecular constituents by hydrolysis, where the hydrolysis may be the result of the activity of chemicals or enzymes. The method can perform actions on molecular constituents that mimic experimental events. For example, fragmentation of an excited molecular component into smaller pieces.

In one embodiment, the invention provides a method of deriving a number of molecular bits k 7 u! resulting from an activity that is assumed to mimic an experimental situation. The pieces may be molecular constituents, such as proteolytic peptides resulting from enzymatic cleavage of a protein, where different assumptions can be made as to the degree of completeness of the enzymatic cleavage. The pieces may be molecular constituents in the form of fragments of molecular constituents, for example fragments of proteolytic peptides.

In one embodiment, the invention provides a method of organizing the masses of molecules or molecular constituents or fragments thereof. Examples of such an organization are given in Figs. 1 and 2., where Fig. 1 shows the number of proteins in a database corresponding to a given proteolytic peptide mass and Fig. 2 shows the grouped distribution of proteolytic peptide masses. Masses that group in this or a similar way will be called a mass distribution peak. Mass cleavage peaks can be recovered for all molecules that contain a limited number of atomic species (eg C, H, N, O, S).

In one embodiment, the invention provides a method that defines the mass ranges in which the frequency of different masses can be determined. The method de-ners f, - as the proportion of masses of molecular constituents or fragments that fall within a mass range i. randomly selected mass distribution peak in the mass range in: P, = FUfwmfß), where F is a function, m, - is a mass range and c denotes experimental and database search constraints. In one embodiment p, - is given by: mass distribution peak.

The denominator in the above expression for p, - denotes the number of mass distribution peaks within the mass range in.

In one embodiment, the invention provides a method of determining the probability of finding a molecular component derived from a particular molecule characterized by ku within a range: Am around a randomly selected molecular stock mass, etc.; = p, -5 (m,., Am), where ó '(m, -, Am) denotes a function that depends on the shape of the mass distribution peak and refers to a mass area. ó '(m, -, Am) can be interpreted as a statistical measure of the number of molecular constituents that can be found within iAm from a randomly selected molecular constituent mass. The mass accuracy may be different before different mass ranges, i.e. in that case denoted by Am in a mass distribution. In one embodiment, δ '(m, -, Am) is determined by integrating a function that describes molecular component mass distributions and standardizing to the total number of molecular component masses in a mass distribution peak.

In one embodiment, ó '(m, -, Am) is determined by direct calculation followed by standardization.

In one embodiment of the invention, a limited number of mass areas are used between m ,,,, - ,, and mm, where -where mass area has an individually defined p, - '.

In one embodiment, probability probes are used to calculate a total probability p (k) for an individual molecule in the database to randomly match k pieces of a total of n masses, where n masses refers to the number of masses in mass data. 10 15 20 w m 12 p (k) = G (p, «Jr, n, ä), where G is a function and c 'denotes experimental and data search constraints.

In an embodiment of the invention, p (k) is given by: w- 2 {»i (1» r (1 pk ,, Z / f,. = K k1l '("1'k1) pl pl k2!' (" 2 "K2) P2 P2 * _ flq 'kg f" n' kr "kl" -k 1 "'Ü_p“)} qqq where q denotes the number of mass ranges, n 1 denotes the number of masses in mass data which is in the range 1, ng denotes the number masses in mass data that are in the range 2 etc, and k 1, with i = 1,2,. .q, denotes the number of matches in mass range i. embodiment of the invention uses a score related to random matching in the process of ranking molecules in a database.

In one embodiment of the invention, the probability p (k) is used in the process of ranking database molecules. An entire database or part of a database is processed and organized to enable the calculation of p (k) for molecules in the database. k denotes the number of matches between the masses of the molecular constituents for each database molecule examined and masses of mass data. The molecules in the database can be known or hypothetical. The molecule or molecules that produce mass data may be known or unknown.

In one embodiment of the invention, the ranking of the molecules is based on the score S (p (k)), where S is a function.

In an embodiment of the invention S (p (k)) - c- <1-zp> - «<¿p, k '10 15 20 25 mr m 13 where c is a constant or a mathematical function. When c = 1, S (p (k)) can be interpreted as the probability that a molecule database would give at least k random correspondences with mass data.

In one embodiment of the invention, the molecule in the database that gives the lowest S @ (k)) gets the highest rank. The molecule that gets the second lowest S (p (k)) gets the second highest rank, and so on. The identification of a molecule or molecules takes place among the molecules that have the highest rank. The highest in rank can be only the highest in rank, but it can also be fl er than the highest in rank, eg the two highest, the three highest, the four highest, the five highest, the ten highest or the 100 maximum. The number of raven-regulated molecules that are considered as identification results can also be determined using a signi fi test.

In one embodiment of the invention, the invention provides a method for generating a frequency function for points for a particular experimental state, where the points are related to random protein identification.

A frequency kt inction is any set of observed values for a studied variable and how many times Each value has been observed. Frequency functions can be in the form of tables of lists, a bar chart, a histogram, a frequency polygon or a continuous curve. Functions derived from frequency distributions can be continuous (probability density function) or discrete (probability mass function). Cumulative benefits of each type of function can also be derived.

In one embodiment, the frequency function is generated for a subset of H members from a database.

In one embodiment, the subgroup is selected based on the values of the cow.

In one embodiment, the subgroup is the fi sequence function generated for molecules ranked by their number of matches.

In one embodiment, the frequency function is f (S), where S is a point. In one embodiment, S is the number of random matches.

In one embodiment, S = k and s17 ~ 2s9 14 k 'H Iu-x H I f (S) = p (k)} - {2 p (k)}, where p (k) has the meaning given above. k = 0 k = 0 Those skilled in the art will recognize that the invention has wide applicability for molecular identification. Although illustrative embodiments of the invention have been described herein, with reference to the accompanying drawings, it is to be understood that the invention is not limited to exactly these embodiments, and that other changes and modifications may be made by one skilled in the art without departing from the scope and spirit of the invention. this invention.

Claims (8)

10 15 20 25 30 51.7 '259 15 Patent claims _ A method for assigning an identity to one or fl your different molecules in a sample by comparing characteristics obtained under certain conditions, said sample having stored characteristics for individual ("stored") molecules, which method is characterized by the steps: a) determining the number, k, of similarities between stored characteristics for said individual molecules and characteristics observed from the sample; b) calculating the probability, p (k), that a certain molecule of said individually stored molecules has characteristics which correspond randomly to the characteristics of the sample; c) assigning an individual score, S @ (k)), for a number of said stored molecules based on the number of matches determined in step (a) and the probability calculated in step (b); d) ranking of each individual stored molecule, which in step (c) has been assigned an individual score, according to this score; and e) assigning an identity to one or fl your molecules, the characteristics of which were obtained under certain conditions, based on the ranking in step (d). The method of claim 1, further characterized in that the determination of the number of matches in step (a) for determining the number of matches in claim 1 is between characteristics of stored molecules assuming that these molecules have been subjected to the same conditions as the molecules in the sample. A method according to claim 1 or 2, further characterized in that said characteristics are masses of the constituents of the stored molecules, which masses are grouped in mass distribution peaks, and that step (b) for calculating a probability in claim 1 comprises the steps : a) determination of the masses and the number, ku, masses that can be generated for the specific ratio for each individual molecule among the stored molecules; b) the fi nition of a total number, q, areas, i, for the masses calculated in step (a); C) determining a proportion, fi, of all the masses calculated in step (a) which are within a range in the fi defined by step (b); _k «__. ¿(mi, Am), m -m- í + l x d) calculation of a probability p! = f, .- where the denominator is the number of mass distribution peaks in the mass range in the fi nied in step (b) above and â (m¿, Am) is a statistical measure of the number of component-selected molecular constituent mass, which means that p, - 'is the probability of finding masses that can be found within: Am from a randomly selected molecular constituent, derived from a certain stored molecule, within a range: Am around a randomly selected constituent mass; e) "determining" the probabilities described in step (d) for all the areas they fi nied in (b); f) determining the number, ni, of masses in mass data falling within each of the q mass regions in the fi nied in step (b); g) determination of the probability Z {_ll! ínpfkl '(l_pfjrnfhllí.p /k¿.(1_p:)nrkz' Mikä, k! :-( n1 - kl) 1 1 kp- (n, -k,) 2 2 ' nq! Ik, lr 75-3, kql-lnq-kqi pq -pql} caused a certain individually stored molecule to match randomly with k pieces out of n masses, where n refers to the number of masses in mass data. A method according to any one of claims 1 to 3, characterized in that said characteristics are masses of the constituents of the stored molecules, g which masses are grouped into mass distribution peaks, and that step (c) of assigning an individual score in claim 1 comprises the step to calculate the score according to S = c - (1 - Z p (k ')), where c is a constant or a function or k' an operator. 10_ 15 517 -259 17 Process according to any one of claims A1 to 4, characterized in that said molecules are biological molecules. Method according to any one of claims 3 to 5, characterized in that said masses have been obtained by mass spectrometry. A method for determining a fun sequence function, f (S), for random molecular identification, based on the method for calculating the probability p (k) according to claim 1, which method is characterized by the steps of: a) defining a subgroup, with H members, of the stored molecules; and b) calculating a frequency function according to 'f (s) = p (k)} p (k)}, where s = k: _ k = 0 k = 0 A method for determining a frequency function, f (S), for random molecular identification, based on the method for calculating the probability p (k) according to claim 3, which method is characterized by the steps: c) the definition of a subgroup, with H members, of the stored molecules, where the members of the subgroup are selected based on their values of ku; and d) calculation of the frequency function according to k, H k, _1 H f (S) = {Z p} - {2 p} »learn S = k1 k = 0 k = 0