A method for automatically identifying pan-modified protein sites based on concentration-dependent relationship of mass spectrum evidence

Abstract

The present application relates to a kind of methods for automatically identifying pan-modified protein sites based on concentration-dependent relationship of mass spectrum evidence.The method obtains the mass spectrum search result and original spectrum information of sample to be analyzed, according to the theoretical mass of small molecule, candidate reaction residue and adduct mass window, candidate covalent adduct PSM is filtered;Further, the spectrum evidence score of candidate PSM and the concentration-dependent score of candidate site are calculated, and the two are combined to form site comprehensive confidence score;According to comprehensive confidence score, output high-confidence, medium-confidence and low-confidence candidate sites, and generate subsequent PRM verification suggestion.The present application can convert manual spectrum inspection basis and concentration response information into standardized, reusable automatic scoring process, reduce subjective difference, improve the efficiency of protein covalent binding site discovery and verification pre-screening.

Description

I. Technical Field

[0001] This invention relates to the fields of mass spectrometry data analysis, bioinformatics, drug-protein interaction analysis, and protein covalent addition site identification, and particularly to a method for automatically identifying panmodified protein sites based on concentration-dependent relationships from mass spectrometry evidence. II. Background Technology

[0002] (a) Current state of technology In drug-protein interaction studies, covalent inhibitor screening, exogenous small molecule adduction modification studies, and drug action mechanism exploration, researchers often need to determine whether a small molecule covalently binds to a protein and further identify its specific binding site. The standard procedure typically includes: searching mass spectrometry data in databases or using open search to screen PSMs that match candidate adduction masses; then, analysts review each MS / MS spectrum to determine the presence of key fragment ions across modification sites, ion continuity on both sides of the modification site, and evidence of precursor mass errors; subsequently, considering the occurrence or signal changes between the blank group, low-concentration group, and high-concentration group, a decision is made on whether to include the candidate site in subsequent PRM targeting validation.

[0003] In existing technologies, candidate site discovery can employ either a restricted database search with preset adductor masses or an open search followed by screening for target mass shifts. For the validation phase, researchers often use software like Skyline to establish PRM or other targeted acquisition methods to reconfirm candidate peptides and characteristic fragment ions. While these methods can identify candidate covalent binding sites to some extent, the transition from candidate screening to site confirmation still heavily relies on manual spectral review and experience-based judgment.

[0004] (ii) Problems with existing technology The existing process has at least the following shortcomings: First, the criteria for judging spectral evidence are not uniform, and different analysts may reach different conclusions as to whether the same candidate PSM is sufficiently credible; Second, comparisons between the blank group, low concentration group, and high concentration group rely heavily on manual observation, lacking reusable and traceable quantitative indicators of concentration dependence; Third, in cases of low abundance samples, weak spectra, a large number of candidate sites, or complex admixture quality, manual verification is costly and prone to missing important candidate sites; Fourth, there is a lack of a unified closed loop between candidate PSM search results, spectral evidence, concentration response, and subsequent PRM validation recommendations, making it difficult to form a standardized project delivery process.

[0005] Therefore, there is an urgent need for a method that can jointly model spectral evidence and sample concentration dependence to automatically identify protein covalent binding sites, score evidence, grade confidence levels, and output validation suggestions, thereby improving the efficiency, consistency, and interpretability of covalent binding site discovery. III. Summary of the Invention

[0006] (a) Purpose of the invention The purpose of this invention is to provide a method for automatically identifying panmodified protein sites based on concentration dependence of mass spectrometry evidence, in order to solve the problems of strong reliance on manual spectrum review, inconsistent scoring standards, lack of automated quantification of concentration response, low efficiency of candidate site verification, and unsystematic output of verification suggestions in the existing covalent site discovery process.

[0007] (II) Technical Solution To achieve the above objectives, this invention provides an automated method for identifying protein covalent binding sites. This method takes mass spectrometry search results, spectral peak tables, sample grouping information, and drug concentration information as input, and outputs candidate PSMs through initial screening, spectral evidence scoring, concentration-dependent scoring, comprehensive confidence grading, and validation suggestions, thus achieving an automated closed loop from candidate site discovery to PRM pre-screening.

[0008] Specifically, the method includes the following steps.

[0009] S1: Data Acquisition. Acquire the mass spectrometry search results and raw spectral information of the sample to be analyzed. The sample includes at least a blank group and one or more drug concentration groups; the mass spectrometry search results include at least candidate peptide sequences, candidate modification quality, candidate modification sites, PSM scores, target / decoy tags, and spectral peak tables, and may further include precursor mass error, fragment ion mass error, retention time, peak area, or precursor intensity.

[0010] S2: Initial screening of candidate PSMs. Based on the input theoretical mass of the small molecule, possible lost groups, candidate reaction residue set, or preset addition mass window, the search results are initially screened to obtain a set of candidate covalent addition PSMs. The candidate addition mass can be derived from the theoretical mass of the small molecule and the mass changes that may occur during the reaction, such as dehydrogenation, dehydration, dehalogenation, and loss of protecting groups.

[0011] S3: Spectral Evidence Score. A spectral evidence score is calculated for each candidate PSM. This score may include the number of b / y ions across modification sites, the peak intensity percentage of key ions across sites, continuity of the ion series before and after modification, precursor quality error, fragment ion quality error, isotopic distribution consistency, control evidence from the unmodified corresponding peptide, and spectra from replicate samples. Figure 1 To the point of being responsive.

[0012] S4: Concentration-dependent score. Calculate a concentration-dependent score for each candidate site. This score may include the degree of absence of the candidate site in the blank group, its frequency of occurrence in the high-concentration group, the increase in the number of spectra with increasing concentration, the trend of peak area or precursor intensity changes, the inter-group differences between the low-concentration and high-concentration groups, and its stability in replicate samples.

[0013] S5: Comprehensive Confidence Grading. The spectral evidence score and concentration dependence score are combined to form a comprehensive confidence score for the candidate locus. The combination method can employ rule-based scoring, weighted summation, machine learning classifiers, Bayesian inference, expert systems, or a combination thereof. Based on the comprehensive confidence score, candidate loci are classified into high-confidence, medium-confidence, and low-confidence categories.

[0014] S6: Validation Recommendation Output. Automatically generates subsequent validation recommendations for high-confidence and medium-confidence candidates, including suggested PRM priority target peptides, recommended monitoring ions, suggested retention time windows, whether to introduce relabeled peptide internal standards, whether to expand the target residue set for re-searching, and whether to perform verification in a full library context.

[0015] Furthermore, the present invention can set a dual search mode: first, improve the sensitivity of candidate discovery in the target protein restriction database, and then verify the candidate sites in the context of the entire database, so as to balance the ability to discover low abundance and control false positives.

[0016] Furthermore, this invention can set prior candidate residues for different chemical reaction types. For example, for thiol-affinity small molecules, the weight of cysteine ​​sites is increased; for nucleophilic substitution or electrophilic addition small molecules, residues such as lysine, arginine, histidine, serine, threonine, and tyrosine that may react are considered, and users are allowed to dynamically adjust the set of candidate residues according to the experimental design.

[0017] (III) Beneficial Effects First, this invention transforms the empirical standards commonly used in manual spectral analysis, such as key fragment ions across modification sites, ion series continuity, and mass error, into calculable indicators, thereby reducing subjective differences among different analysts.

[0018] Second, this invention incorporates the concentration response information between the blank group, the low concentration group, and the high concentration group into the site reliability score, which can enhance the ability to identify real drug adduct sites and reduce false positives caused by accidental matching or background noise.

[0019] Third, this invention forms an integrated process from initial screening of candidate PSMs, spectral evidence scoring, concentration-dependent scoring, confidence level classification to PRM verification recommendation output, which can reduce the burden of manual review and improve project delivery efficiency.

[0020] Fourth, this invention is applicable to both the exploration of drug action mechanisms in scientific research scenarios and the evaluation of covalent binding of candidate drugs in enterprise R&D, and has high application value and reusability. IV. Description of the attached drawings

[0021] Figure 1 This is a schematic diagram of the overall process of the automatic identification method for protein covalent binding sites of the present invention.

[0022] Figure 2 This is a schematic diagram of the spectral evidence scoring method in this invention.

[0023] Figure 3 This is a schematic diagram of the concentration-dependent scoring in this invention. V. Detailed Implementation Methods

[0024] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. All equivalent substitutions, modifications, or combinations made based on the inventive concept should fall within the scope of protection of the present invention.

[0025] Implementation Method 1: Overall Process like Figure 1 As shown, in this embodiment, the mass spectrometry search results, spectral peak tables, and sample concentration information of the blank group and multiple drug concentration groups are first input; then, candidate covalent adduct PSMs are screened based on the theoretical mass of the target small molecule and the candidate reaction residue set; then, the spectral evidence score and concentration dependence score are calculated separately; then, the two are combined to form a comprehensive confidence score, and the confidence is classified as high, medium, and low according to the threshold or model output results; finally, the candidate site table and PRM validation suggestions are output.

[0026] Implementation Method 2: Initial Screening of Candidate PSMs In one specific embodiment, the samples to be analyzed include a blank group, a low-concentration group, a medium-concentration group, and a high-concentration group. First, candidate peptide sequences, candidate modification qualities, candidate modification residues, PSM scores, and spectral indices are retrieved from database searches, open searches, or restricted search results. Then, a candidate adduct quality list is generated based on the theoretical quality of the small molecule and possible changes in reaction quality, and a quality matching window is set, such as a quality deviation window in ppm or Da. When a candidate modification quality in the search results falls within the candidate adduct quality window, and the candidate modification residue belongs to a preset set of candidate reaction residues, the PSM is included in the candidate covalent adduct PSM set.

[0027] Implementation Method 3: Spectral Evidence Scoring like Figure 2As shown, for each candidate PSM, this embodiment matches theoretical fragment ions with the measured MS / MS peak tables to determine whether there are key fragment ions supporting localization on both sides of the modification site. If there are a large number of b / y ions across the modification site, the key ion intensity is high, the ion series continuity is good, and the precursor mass error and fragment ion mass error are within a preset range, then the candidate PSM obtains a high spectral evidence score. Conversely, if key cross-site ions are missing, the spectral interpretation ratio is low, or the mass error is large, then the spectral evidence score is reduced.

[0028] Spectral evidence scores can be obtained through rule-based scoring. For example, a certain score is given for each key ion appearing across modification sites; the higher the proportion of the key ion peak intensity to the total interpreted peak intensity, the higher the score; the more complete the series of consecutive ions on both sides of the modification site, the higher the score; and a penalty is imposed when the quality error exceeds a threshold. Alternatively, machine learning models can be used to classify or regress candidate PSMs using the above features as input.

[0029] Implementation Method 4: Concentration-dependent scoring like Figure 3 As shown, for the same candidate site, this embodiment statistically analyzes the frequency of occurrence, number of spectra, peak area, or precursor intensity of the site in the blank group, low concentration group, medium concentration group, and high concentration group. If the candidate site is absent or significantly low in the blank group, but appears in the administered group, and shows an increasing trend in the number of spectra, enhanced peak area, or increased precursor intensity with increasing administered concentration, then the candidate site obtains a high concentration-dependent score.

[0030] Concentration-dependent scores can be obtained using monotonicity tests, trend regression, inter-group comparisons, order statistics, or rule-based scoring methods. For example, a positive score can be given for missing data in the blank group; a positive score can be given for high frequency of occurrence in the high concentration group; a trend score can be given for the number of spectra or peak area that increases with increasing concentration; and a penalty can be imposed on candidate sites in the blank group that have obvious signals or inconsistent concentration trends.

[0031] Implementation Method 5: Comprehensive Confidence Level Grading and Output After obtaining the spectral evidence score and concentration dependence score, this implementation method combines the two to form a comprehensive confidence score. The combination method can be a weighted summation, for example, comprehensive confidence score = w1 × spectral evidence score + w2 × concentration dependence score, where w1 and w2 are preset weights; it can also be implemented using a classification model, Bayesian inference, or expert rules.

[0032] Based on the overall confidence score, candidate sites are categorized into high-confidence, medium-confidence, and low-confidence. High-confidence candidates can be directly listed as priority confirmation sites; medium-confidence candidates are listed as sites recommended for manual review; and low-confidence candidates can be marked as low-priority, downgraded, or eliminated. The output should include at least the protein name, peptide sequence, candidate modification site, candidate adduct quality, corresponding sample or concentration group, spectral evidence score, concentration-dependent score, overall confidence score, grading result, and validation recommendations.

[0033] Implementation Method Six: PRM Validation Recommendation Generation For high-confidence and medium-confidence candidate sites, this implementation further generates PRM verification recommendations. Specifically, based on the spectral evidence and fragment ion matching of candidate peptides, fragment ions with high intensity and site localization significance are selected as recommended monitoring ions; the PRM retention time window is set according to the retention time information in the DDA or DIA data; for high-value but low-abundance candidate sites, it is recommended to introduce relabeled peptide internal standards; for cases where the range of candidate residues is narrow or the interpretation of mass shift is not unique, it is recommended to expand the target residue set or use full library background verification.

[0034] Implementation Method Seven: Alternative Implementation Method As alternative implementations, spectral evidence scores can be achieved using rule-based scoring, machine learning classifiers, Bayesian inference, or expert systems; concentration dependence can be achieved using monotonicity tests, trend regression, inter-group comparisons, order statistics, or other statistical models; and the candidate residue set can be manually input by the user, automatically recommended by chemical rules, or predicted by a trained model. As long as the core remains the same—combining spectral evidence with sample concentration dependence to identify and confirm protein covalent binding sites and outputting grading results and validation suggestions—it can be considered an alternative implementation of this invention.

[0035] Implementation Method 8: Example of Output Results Target protein A ACDEK*LMN K123 +X Da high high High reliability Target protein B FGHC*TVR C45 +X Da middle high Zhongxin Target protein C LMNR*QST R88 +X Da Low middle Low reliability

Claims

1. A method for automatically identifying panmodified protein sites based on concentration dependence evidence from mass spectrometry, characterized in that, The process includes the following steps: obtaining mass spectrometry search results and raw spectral information of the sample to be analyzed, wherein the sample to be analyzed includes at least a blank group and one or more drug concentration groups; screening a set of candidate covalently added PSMs from the mass spectrometry search results based on the input small molecule theoretical mass, candidate reaction residue set, and / or preset addition mass window; calculating a spectral evidence score for each candidate PSM in the candidate covalently added PSM set; calculating a concentration-dependent score for the occurrence and signal changes of the same candidate modification site in different samples or different concentration groups; combining the spectral evidence score with the concentration-dependent score to obtain a comprehensive confidence score for the candidate modification site; classifying the candidate modification sites according to the comprehensive confidence score, and outputting protein covalent binding site identification results and verification suggestions.

2. The method according to claim 1, characterized in that, The mass spectrometry search results include one or more of the following: candidate peptide sequence, candidate modification quality, candidate modification site, PSM score, target / decoy tag, precursor quality error, retention time, peak area, or precursor intensity; the raw spectrum information includes an MS / MS spectrum peak table or a peak list converted from the raw mass spectrometry data.

3. The method according to claim 1, characterized in that, The steps for screening candidate covalent adduct PSM sets include: generating one or more candidate adduct masses based on the theoretical mass of the small molecule and possible mass changes due to dehydrogenation, dehydration, dehalogenation, loss of protecting groups, or other reactions; matching the candidate adduct masses with candidate modification masses in the mass spectrometry search results; and retaining or downweighting the candidate PSMs according to the candidate reaction residue set.

4. The method according to claim 1, characterized in that, The spectral evidence score includes one or more of the following indicators: the number of b or y ions across the modification site, the peak intensity ratio of key fragment ions across the modification site, the continuity of the ion series on both sides of the modification site, precursor mass error, fragment ion mass error, isotope distribution consistency, control evidence of the corresponding unmodified peptide, and spectral consistency of the same candidate site in repeated samples.

5. The method according to claim 1, characterized in that, The concentration-dependent score includes one or more of the following indicators: the degree of absence of candidate sites in the blank group, the frequency of occurrence in the administered group, the increase in the number of spectra with increasing administered concentration, the trend of peak area or precursor intensity with increasing administered concentration, the inter-group difference between the low-concentration group and the high-concentration group, and the stability of candidate sites in biological or technical replication.

6. The method according to claim 1, characterized in that, The overall confidence score is obtained through rule scoring, weighted summation, machine learning classifiers, Bayesian inference, expert systems, or a combination thereof, wherein the spectral evidence score and concentration dependence score are each independent sources of evidence participating in the overall score.

7. The method according to claim 1, characterized in that, The grading includes classifying candidate modification sites into high-confidence candidates, medium-confidence candidates, and low-confidence candidates; High-confidence candidate outputs are priority confirmation sites, medium-confidence candidate outputs are sites recommended for manual review, and low-confidence candidates are downgraded, marked, or removed.

8. The method according to claim 1, characterized in that, The validation recommendations include one or more of the following: PRM priority target peptides, recommended monitoring of fragment ions, recommended retention time windows, whether to introduce relabeled peptide internal standards, whether to expand the candidate reaction residue set for re-search, and whether to use full library background verification.

9. The method according to claim 1, characterized in that, The method also includes a dual search mode: first, candidate discovery is performed in the target protein restriction database to improve the sensitivity of finding low-abundance candidate covalent sites; then, candidate sites are verified in the context of the entire database to reduce false positives caused by accidental matching or background noise.

10. The method according to claim 1, characterized in that, The set of candidate reactive residues is dynamically determined based on the chemical reaction type of the small molecule; for thiol-affinity small molecules, the weight of cysteine ​​sites is increased; for nucleophilic substitution or electrophilic addition small molecules, lysine, arginine, histidine, serine, threonine, tyrosine, or other potentially reactive amino acid residues are considered.

Images