Motivation
For as long as computers have been used to predict the function of a protein or gene, the general approach has been this: search a database for similar sequences, then leverage what is known about those matches to annotate the molecule of interest. Whether through sequence alignment, structural comparison, or protein language models, the core strategy remains unchanged: base the annotation of a molecule on its detected similarity to some previously annotated sequence.
Despite various technological advances, the number of genes and protein sequences without useful functional annotation remains stubbornly high at 30-50%, even in well-characterized organisms, and will not be easily addressed by traditional or emerging homology- or structural-similarity based approaches.
This session addresses a grand challenge in computational biology: moving beyond homology- and similarity-based methods for function annotation to understand function from first principles.
Building on PSB 2026
Our session builds on the discussions and excitement from PSB 2026, where we posed the question: "What is the frontier for function? What is the greatest barrier to determination of function?" The discussion revolved around the limits of homology, and the community expressed strong interest in methods that could transcend or bypass homology to provide functional annotations for unknown proteins and genes [1].
The topic is particularly relevant now because AI methods have driven the limits of detection of homology down into the 'twilight zone' [1]. It's time to explore what lies beyond.
Topics of Interest
We seek papers that push the boundaries of what can be achieved beyond the homology paradigm [1]:
Protein/gene function prediction approaches that don't rely exclusively on similarity to previously annotated proteins
Direct prediction of function from AI models trained on sequence or structural features
Inference based on similarity of binding or enzymatic active sites
Using protein structure to understand and predict function
Integration of multi-omic or other contextual data for function prediction
Harnessing evolution and/or gene context for protein function prediction
Creative integration of experimental data and biological network information
Approaches to interpretable AI focused on molecular function
Approaches for re-defining functional ontologies and interpreting free text annotations
Approaches addressing difficult function problems in viruses, non-model organisms, and hard-to-predict protein families
Entirely novel frameworks for linking protein characteristics to biological role
Identification of novel protein/gene functionality for biomedical or other applications
Session Organizers
Jason McDermott
Pacific Northwest National Laboratory
Yana Bromberg
Emory University
Hannah Carter
University of California San Diego
Travis Wheeler
University of Arizona
Submission Information
Paper Format and Submission Portal
Please see the PSB paper format template and instructions at http://psb.stanford.edu/psb-online/psb-submit
Cover Letter Requirements
Each paper must be accompanied by a cover letter as the first page of your submission. The cover letter must state:
- The email address of the corresponding author
- Indicate that your submission is directed to the "Biological Molecular Function: Beyond Homology" session
- The submitted paper contains original, unpublished results, and is not currently under consideration elsewhere
- All co-authors concur with the contents of the paper