AIomics: Exploring More of the Proteome Using Mass Spectral Libraries Extended by Artificial Intelligence

Abstract

The unbounded permutations of biological molecules, including proteins and their constituent peptides, present a dilemma in identifying the components of complex biosamples. Sequence search algorithms used to identify peptide spectra can be expanded to cover larger classes of molecules, including more modifications, isoforms, and atypical cleavage, but at the cost of false positives or false negatives due to the simplified spectra they compute from sequence records. Spectral library searching can help solve this issue by precisely matching experimental spectra to library spectra with excellent sensitivity and specificity. However, compiling spectral libraries that span entire proteomes is pragmatically difficult. Neural networks that predict complete spectra containing a full range of annotated and unannotated ions can be used to replace these simplified spectra with libraries of fully predicted spectra, including modified peptides. Using such a network, we created predicted spectral libraries that were used to rescore matches from a sequence search done over a large search space, including a large number of modifications. Rescoring improved the separation of true and false hits by 82%, yielding an 8% increase in peptide identifications, including a 21% increase in nonspecifically cleaved peptides and a 17% increase in phosphopeptides.

Rescoring Peptide Spectrum Matches: Boosting Proteomics Performance by Integrating Peptide Property Predictors Into Peptide Identification

“…While SLSEs inherently benefit from comprehensive predicted libraries with reference-like quality ( 115 ), one main challenge still remains: The generated libraries may become prohibitively large, affecting storage, sharing, and utilization. In a recent study, a predicted spectral library was used to showcase the performance of Mistle in metaproteomics ( 114 ).…”

“…While the authors demonstrated that Mistle works well in this use case, extrapolating such a library to all possible HLA class I peptides from a canonical human proteome would require generating, storing, and handling an msp file with more than 250 GB in size, storing upwards of 150 M spectra for ∼45 M unmodified peptide sequences. This size would increase even further when considering possible PTMs ( 116 ), mutations ( 115 ), or even cryptic HLA peptides ( 32 ). Furthermore, such libraries will likely have to be generated for every mass spectrometer and various instrument settings separately in order to maintain the high specificity of the library.…”