Biology subjects

Lapin, J.

Publications and source records attributed to Lapin, J..

5 recordsLinked to original sources

bioRxiv2025

Systematic evaluation of peptide property predictors with explainable AI technique SHAP

Deep learning models are often characterized as black boxes because their layers of various mathematical transformations and activation functions are practically uninterpretable and have little meaning to the user. However, explainable AI methods exist to attribute model output predictions to its inputs. Shapley additive explanations (SHAP) is one such method that directly quantifies the inputs contributions and qualitatively addresses the question of why a model makes a particular prediction. SHAP generally is hampered by its computational cost, which scales very poorly with large inputs. Therefore peptide property predictors that take as input amino acid sequences, roughly between the lengths of 10-30 amino acids, represent ideal systems for applying SHAP. In applying SHAP to models that predict retention time, collisional cross section, peptide flyability, and fragment intensity, we obtain the relative influence that each amino acid has on predicting each property, and furthermore can rationalize the values from the perspective of the amino acids chemistry. Simply correlating the average shapley values per amino acid type over an entire validation dataset has yielded high correlation to published amino acid indices that are highly related to the property being predicted. For instance, our average shapley values for retention time had a 0.973 Pearson correlation with experimentally measured amino acid retention indices at pH 2. In applying SHAP on the Prosit fragment intensity prediction model, there is strong agreement with the mobile proton model, specifically demonstrating the effect of basic amino acids, the proline effect in charge-remote fragmentation, and positively charged amino acids in charge-directed fragmentation. We also use SHAP in a targeted experiment to demonstrate the Pathways in competition behavior of the model, and reveal a very discrete decision process based on basic residues and the fragment charge. While SHAP in this work was applied to models of well understood properties/systems, there is great potential to explain less studied areas of peptide chemistry to provide insights into their mechanisms.

bioinformatics↗

bioRxiv2025

Integrating Alternative Fragmentation Techniques into Standard LC-MS Workflows Using a Single Deep Learning Model Enhances Proteome Coverage

We built and characterised a mass spectrometer capable of performing CID (both beam type and resonant type), UVPD, EID and ECD in an automated fashion during an LCMS type experiment. We exploited this ability to generate large datasets through multienzyme deep proteomics experiments for characterisation of these activation techniques. As a further step, motivated by the complexity generated by these dissociation techniques, we developed a single Prosit deep learning model for fragment ion intensity prediction covering all of these techniques. The generated model has been made publicly available and has been utilised in FragPipe within its MSBooster module. Rescoring allowed both data-dependent acquisition (DDA) and data-independent acquisition (DIA) to achieve on average more than 10% increase in protein identifications across all dissociation techniques and enzymatic digestions. We demonstrate that these alternative fragmentation approaches can now be used within standard data analysis pipelines and can produce data competitive to CID in terms of eficiency, but in the cases of EID and UVPD with far richer and more comprehensive spectra.

systems biology↗

bioRxiv2025

Pairwise Attention: Leveraging Mass Differences to Enhance De Novo Sequencing of Mass Spectra

A fundamental challenge in mass spectrometry-based proteomics is determining which peptide generated a given MS2 spectrum. Peptide sequencing typically relies on matching spectra against a known sequence database, which in some applications is not available. Deep learning-based de novo sequencing can address this limitation by directly predicting peptide sequences from MS2 data. We have seen the application of the transformer architecture to de novo sequencing produce state-of-the-art results on the so-called nine-species benchmark. In this study, we propose an improved transformer encoder inspired by the heuristics used in the manual interpretation of spectra. We modify the attention mechanism with a learned bias based on pairwise mass differences, termed Pairwise Attention (PA). Adding PA improves average peptide precision at 100% coverage by 12.7% (5.9 percentage points) over our base transformer on the original nine-species benchmark. We have also achieved a 7.4% increase over the previously published model Casanovo. Our MS2 encoding strategy is largely orthogonal to other transformer-based models encoding MS2 spectra, enabling straightforward integration into existing deep-learning approaches. Our results show that integrating domain-specific knowledge into transformers boosts de novo sequencing performance.

bioinformatics↗

bioRxiv2024

Prosit-XL: enhanced cross-linked peptide identification by accurate fragment intensity prediction to study protein-protein interactions and protein structures

It has been shown that integrating peptide property predictions such as fragment intensity into the scoring process of peptide spectrum match can greatly increase the number of confidently identified peptides compared to using traditional scoring methods. Here, we introduce Prosit-XL, a robust and accurate fragment intensity predictor covering the cleavable (DSSO/DSBU) and non-cleavable cross-linkers (DSS/BS3), achieving high accuracy on various holdout sets with consistent performance on external datasets without fine-tuning. Due to the complex nature of false positives in XL-MS, a novel approach to data-driven rescoring was developed that benefits from Prosit-XLs predictions while limiting the overestimation of the false discovery rate (FDR). We first evaluated this approach using two ground truth datasets that demonstrate the accurate and precise FDR estimation. Second, we applied Prosit-XL on a proteome-scale dataset, demonstrating an up to [~]3.4-fold improvement in PPI discovery compared to classic approaches. Finally, Prosit-XL was used to increase the coverage and depth of a spatially resolved interactome map of intact human cytomegalovirus virions, leading to the discovery of previously unobserved interactions between human and cytomegalovirus proteins.

bioinformatics↗

bioRxiv2023

UniSpec: A Deep Learning Approach for Predicting Energy-Sensitive Peptide Tandem Mass Spectra and Generating Proteomics-Wide In-Silico Spectral Libraries

In this report, we present UniSpec, an attention-based deep neural network designed to predict complete collision-induced fragmentation of tryptic peptides, aimed at enhancing peptide and protein identification in shotgun proteomics studies. We preprocessed spectral data from peptide tandem mass spectral libraries, compiled by the National Institute of Standards and Technology (NIST), utilizing a data approach tailored for model development, resulting in high-quality, energy-consistent spectral datasets. By analyzing all the annotated fragment ions present in these libraries, we constructed an extensive peptide fragment dictionary containing 7919 isotopic ions from sequence ions, neutral loss, internal, iminium, and amino acid fragment ions. The streamlined dictionary-based spectral training data enables UniSpec to efficiently learn the complex intensity patterns of various product ions, resulting in reliable spectral predictions for a wide range of unmodified and modified peptides. We evaluated the models accuracy by comparing its performance across training and testing data, considering diverse peptide characteristics like peptide classes, charge states, and sequence lengths. Our model attained a median cosine similarity score of 0.951 and 0.923 on the training and test data respectively. Contrary to existing deep learning models that often overlook a substantial part of peptide tandem mass spectra beyond the sequence b and y ion series, UniSpec can predict up to 75% of all measured fragment intensities (including unknown signals) in the raw experimental spectra. This represents a marked advancement from the 43.5% coverage achieved solely by b and y sequence ions in the NIST library spectra. For the evaluation of our models practical utility in predicting proteome-wide in-silico spectral libraries, we executed a benchmark test using a dataset of HeLa cells. UniSpec displayed a significant overlap of peptide identifications with the widely used search engine MS-GF+ and the NIST experimental spectral library, demonstrating its robust performance as a standalone peptide identification tool.

bioinformatics↗