AlphaPeptDeep: A modular deep learning framework to predict peptide properties for proteomics

Abstract

Machine learning and in particular deep learning (DL) are increasingly important in mass spectrometry (MS)-based proteomics. Recent DL models can predict the retention time, ion mobility and fragment intensities of a peptide just from the amino acid sequence with good accuracy. However, DL is a very rapidly developing field with new neural network architectures frequently appearing, which are challenging to incorporate for proteomics researchers. Here we introduce AlphaPeptDeep, a modular Python framework built on the PyTorch DL library that learns and predicts the properties of peptides (https://github.com/MannLabs/alphapeptdeep). It features a model shop that enables non-specialists to create models in just a few lines of code. AlphaPeptDeep represents post-translational modifications in a generic manner, even if only the chemical composition is known. Extensive use of transfer learning obviates the need for large data sets to refine models for particular experimental conditions. The AlphaPeptDeep models for predicting retention time, collisional cross sections and fragment intensities are at least on par with existing tools. Additional sequence-based properties can also be predicted by AlphaPeptDeep, as demonstrated with a novel HLA peptide prediction model to improve HLA peptide identification for data-independent acquisition.

MSBooster: Improving Peptide Identification Rates using Deep Learning-Based Features

“…Support for more DL models may also benefit HLA rescoring, where models such as Prosit have been trained on nonspecific peptides [47]. Finally, transfer learning and fine-tuning implemented in pDeep3 [40] and AlphaPeptDeep [41] may help to analyze MS/MS spectra acquired using different fragmentation mechanisms, or when identifying peptides containing rare PTMs.…”

“…More recently, however, a wave of deep learning (DL) models have been trained to predict the physicochemical properties of peptides and MS/MS spectra [37][38][39][40][41][42]. By training on millions of available peptides, these models can learn general rules to make accurate predictions for new peptides, assuming they are not vastly different from those on which the models were trained.…”

The potential of plasma HLA peptides beyond neoepitopes

“…Three different strategies are commonly used: experimental libraries, typically acquired by DDA; pseudospectra-based libraries extracted by directDIA as introduced by DIA-Umpire 16 and implemented in Spectronaut; and libraries in which fragment intensities are predicted by deep learning 14,17 . In connection with the latter approach, we recently introduced a deep learning based framework called AlphaPeptDeep, which predicts spectral libraries tailored for different MS platforms, only based on a database file of the proteome in FASTA format or just a peptide list as input 18 . It contains the PeptDeep-HLA model which makes use of the inherent similarity of immunopeptides present within one person based on their HLA type.…”

“…As part of our workflow, we have also implemented Data Independent Acquisition (DIA) to expand the depth of the immunopeptidomic data. To tackle the challenge of creating a suitable search space for immunopeptidomics, we employed personalized HLA peptide libraries 18 . This considerably reduces the number of potential 9mers to 12mers in a human FASTA to be searched, increasing the number of significant identifications.…”

Synchro-PASEF allows precursor-specific fragment ion extraction and interference removal in data-independent acquisition

“…Frequently used software tools such as DIA-NN (2,3) or Spectronaut (4) match peptides from a library into each of the DIA runs. These libraries were traditionally acquired experimentally with deep data dependent acquisition (DDA)-based measurements of the proteome of interest, but are now often generated directly from the DIA data (5) or in silico from the entire proteome using deep learning (6)(7)(8)(9)(10)(11). Given these advantages very deep and quantitatively accurate data sets can now routinely be generated by DIA.…”