Sano’s deepFRI2 at ISMB 2026
Postdoctoral researcher Paweł Szczerbiak presented Sano's interpretable protein-function prediction model to more than 2,000 computational biologists in Washington, D.C
Sano was represented at an important international conference in computational biology this July, as Paweł Szczerbiak, a postdoctoral researcher in Sano’s Structural and Functional Genomics Group, took the stage at the 34th Conference on Intelligent Systems for Molecular Biology (ISMB 2026).
Held from 12 to 16 July at the Washington Hilton in Washington, D.C., ISMB is the flagship meeting of the International Society for Computational Biology (ISCB). This year the conference returned to the city where ISCB and ISMB were founded back in 1993, drawing more than 2,000 computational life scientists from around the world. Paweł Szczerbiak presented both a poster and a flash talk titled “Interpretable protein function predictions with deepFRI2.”
deepFRI2 is an upgraded framework for predicting the biological function of proteins directly from their sequence and structure. What sets it apart is its focus on interpretability: instead of behaving like a “black box,” the model points to the specific regions of a protein that drive a given functional prediction. Remarkably, it reaches state-of-the-art accuracy using only a few million parameters, making it both efficient and scalable — qualities that matter increasingly as researchers apply artificial intelligence to a vast and fast-growing catalogue of protein structures.
For Szczerbiak, the conference was as much about listening as presenting. He singled out the keynote by Marinka Zitnik of Harvard Medical School — an “extraordinary” talk on AI-driven discovery in medicine — as a personal highlight, noting how sharply the field has advanced in just one year, in step with the rapid progress of language models. Meetings of this scale, he added, are an invaluable chance to map the frontiers of bioinformatics, exchange ideas, and stay current with the trends shaping the discipline.
The deepFRI2 project was also the subject of a recent Sano Seminar, “Interpretable protein function predictions with deepFRI2,” for anyone who would like to explore the work in more depth (https:/interpretable-protein-function-predictions-with-deepfri2).
Sano’s presence at ISMB 2026 underscores the institute’s growing role in the international computational medicine community — and its commitment to developing AI tools that are not only powerful, but also transparent and trustworthy.
