Robust Accurate Identification of peptides from tandem mass spectrometry data

National Library Of Medicine

Investigators

Linked publications, trials & patents

Abstract

Assigning statistical significance accurately has become increasingly important as metadata of many types, often assembled in hierarchies, are constructed and combined for further biological analyses. Statistical inaccuracy of metadata at any level may propagate to downstream analyses, undermining the validity of scientific conclusions thus drawn. From the perspective of mass spectrometry-based proteomics, this implies that accurate statistics must be obtained in peptide identification, then built on if one can hope to have protein identification method(s) with accurate statistical significance assignment. However, although heavily concentrated upon and studied, the statistical accuracy of peptide/protein identification remains challenging. There are many peptide identification methods using database searches and assigning the E-value to peptide hits; however, the E-values reported by different methods do not agree with each other and few of them, if any, agree with the textbook definition of the E-value. This obviously hinders the feasibility of combining search results from different methods, particularly if one wishes to combine methods with user-assigned weights. When prior knowledge is available, it is often desirable to weight search methods differently before combining their search results. In our earlier publications, we have developed peptide identifications methods with accurate statistical significance assignment founded on the extension of the central limit theorem, and all possible peptide statistics; we have provided a way to combine search results democratically in one of our earlier publications. When different weights are present, an instability issue occurs if some of the weights are nearly degenerate; we have devised a mathematical framework to completely eliminate this instability. In 2015, we published a protein identification method that combines weighted P-values of evidence peptides. This new method solves the long-standing problem of precise type-I error control in protein identification. In addition, it also reports correctly the proportion of false discoveries, indication of accurate type-II error control. In 2016, we published a new peptide significance assignment method based on the extreme value statistics. The motivation of this work is to provide accurate peptide identification confidence for methods that use scoring functions that cannot be expressed as a sum of independent contributions. In the past years, we also worked on a large cross-institute project, involving scientists in NHLBI, NCI and Clinical Center, in pathogen identifications using mass spectrometry. The fundamental idea is to use each pathogen's peptidome to represent that pathogen. Through mass spectrometry analysis, if the statistical significance assignment is accurate, one will be able to correctly rank the species/genus according to their peptidome similarity compared with the peptides identified. Again, we have to weigh the evidence peptides associated with a given species/genus as one peptide often maps to multiple species/genus. Between 2017-2020, we have finished the first three phases of the study: namely, identification of a single microbe, simultaneous identifications of multiple microbes, and simultaneous identification of multiple microbes along with their protein biomass estimates. The third phase is made possible by introducing several new ideas into the analysis pipeline: taxon priors, ownership, participation ratios, and degree of independence. With these quantities properly computed, one is able to estimate the taxa protein biomass contributions, the number of taxa to keep in a taxa cluster and to split off a sufficiently independent taxon off a cluster. The last point is important in alleviating the effect due to aggressive clustering. Results from the first three phases were all published in the Journal of American Society of Mass Spectrometry. In 2021 and 2022, we focused on making our tools more accessible and on establishing international collaborations. In addition to writing and keeping up the graphic user interface program for our microbial identification and biomass estimate method, we have also collaborated with the Camadro group of University of Paris and the CCUG group at the University of Gothenburg in Sweden. Our collaboration with the University of Gothenburg focused on the study of antibiotic resistance of pathogens. Within the framework built in phase three, for protein identifications, we added another database containing bacterial proteins that are known to be related to microbe's antibiotic resistance. This new addition can be very helpful in identification of potential antibiotic resistance and thus avoid the use of ineffective antibiotics in treatment. This new feature and the associated new results were published in the Journal of American Society of Mass Spectrometry in 2021 and 2022. This year we have focused on implementing new features to our pathogen identification graphical user interface (GUI) based on user requests and comments in order to enhance usefulness and ease of use. The new features include improved tools for workflow, reproducibility and record keeping, data analysis and visualization, and new or enhanced tools for computing and studying peptide fragmentation, peptide/protein isotopic distributions, and lowest common ancestor algorithm for proteotyping bio-marker design. We have submitted a manuscript, currently under review, describing our pathogen identification GUI to the Journal of Computational Biology. Sensitivity and specificity of peptide identification could potentially be improved if fragmentation intensity profiles of candidate peptides were known. To this end, several deep learning-based methods have recently been proposed. However, an independent, systematic assessment of accuracy of predictions and a detailed analysis of usability of these methods for various types of experimental data is lacking. Therefore, this year we also performed a comprehensive performance comparison between several state-of-the-art prediction methods using a large array of diverse, publicly available mass spectrometry data.