Broad functional profiling of fission yeast proteins using phenomics and machine learning

Area of Science:

• * Molecular Biology
• * Systems Biology
• * Computational Biology

Background:

• * A significant number of proteins lack functional characterization, hindering biological research.
• * Understanding protein function is crucial for advancing various biological disciplines.

Purpose of the Study:

• * To apply phenomics and machine learning to systematically characterize protein functions in *Schizosaccharomyces pombe*.
• * To identify novel functions for uncharacterized and 'priority unstudied' proteins, including those conserved in humans.

Main Methods:

• * Assayed colony-growth phenotypes for 3509 non-essential gene deletion mutants across 131 diverse conditions.
• * Constructed phenotype-correlation networks to infer protein functions via 'guilt by association'.
• * Employed machine learning (NET-FF) integrating protein-network and homology data to predict Gene Ontology (GO) terms.

Main Results:

• * Phenotypic data was obtained for 3492 mutants, revealing functional clues for 124 'priority unstudied' proteins.
• * Over 900 proteins were newly linked to oxidative stress resistance.
• * Generated 56,594 high-scoring GO predictions, with 1675 novel predictions for 783 genes, including 47 for 23 priority unstudied proteins.

Conclusions:

• * Integrated phenomics and machine learning provide a powerful resource for discovering protein functions.
• * Experimental validation confirmed novel protein roles in cellular aging.
• * The study significantly expands the functional annotation of the *S. pombe* proteome and identifies key targets for future research.