The Age of Data-Driven Proteomics: How Machine Learning Enables Novel Workflows.

Area of Science:

• Proteomics
• Computational Biology
• Mass Spectrometry

Background:

• Challenging proteomics workflows such as metaproteomics, proteogenomics, and data-independent acquisition (DIA) face significant ambiguity during peptide and protein identification.
• This ambiguity arises from expanded search spaces or inherently complex data, hindering accurate analysis in liquid chromatography-mass spectrometry (LC-MS) experiments.

Purpose of the Study:

• To highlight the potential of machine learning (ML)-based predictive models in mitigating identification ambiguity within advanced proteomics techniques.
• To discuss recent advancements in ML for proteomics and identify areas requiring further development and validation.

Main Methods:

• Review of existing machine learning and deep learning models applied to various aspects of LC-MS experiments.
• Analysis of the challenges and limitations in integrating predictive models into complex proteomics workflows.

Main Results:

• ML models have demonstrated capability in predicting multiple parameters of LC-MS experiments, offering solutions to identification ambiguity.
• Despite progress, the integration of these predictive models into routine challenging workflows remains limited.

Conclusions:

• ML holds significant promise for enhancing the accuracy and efficiency of complex proteomics identifications.
• Continued research in ML model development, validation, and integration is crucial for advancing the field of proteomics.