NovoRank:\nRefinement for <i>De Novo</i> Peptide\nSequencing Based on Spectral Clustering and Deep Learning

<i>De novo</i> peptide sequencing is a valuable\ntechnique\nin mass-spectrometry-based proteomics, as it deduces peptide sequences\ndirectly from tandem mass spectra without relying on sequence databases.\nThis database-independent method, however, relies solely on imperfect\nscoring functions that often lead to erroneous peptide identifications.\nTo boost correct identification, we present NovoRank, a postprocessing\ntool that employs spectral clustering and machine learning to assign\nmore plausible peptide sequences to spectra. Prior to <i>de novo</i> peptide sequencing, spectral clustering is applied to group similar\nspectra under the assumption that they originated from the same peptide\nspecies. NovoRank then employs a deep learning model, incorporating\nboth cluster-derived proteomic features and individual spectrum characteristics,\nto rerank the candidate peptides produced by <i>de novo</i> peptide sequencing. Our results show that NovoRank significantly\nenhances the performance of various <i>de novo</i> peptide\nsequencing tools, increasing both recall and precision by 0.020 to\n0.080 at the peptide-spectrum match (PSM) level. Notably, NovoRank\nachieves a recall as high as 0.830 for Casanovo at the PSM level.\nThe source code of NovoRank is freely available at https://github.com/HanyangBISLab/NovoRank and is licensed under Creative Commons Attribution-NonCommercial-ShareAlike\n4.0 International.