Skip to content

S3 — Systems Biology and Network Analysis

Full session title: Systems Biology and Network Analysis: From Multi-omics Integration to Biological Mechanisms Papers: 6 (pp. 755–848) Session introduction: Bebek G. et al., pp. 752–754 · PDF · PMID 41758182

Organizers:

  • Gurkan Bebek — Center for Proteomics and Bioinformatics, Case Western Reserve University
  • Onur Mutlu — Information Technology and Electrical Engineering, ETH Zürich
  • Iman Hajirasouliha — Systems and Computational Biomedicine, Weill Cornell Medicine
  • Joshua Welch — Computational Medicine and Bioinformatics, University of Michigan
  • Serguei Pakhomov — Pharmaceutical Care & Health Systems, University of Minnesota

Invited talk: Wenyi Wang (Mon Jan 5, 9:50 a.m.) — affiliation not printed in the schedule.

Reading this session

The smallest peer-reviewed session, and the most methodologically classical — graphs, diffusion, random walks, knowledge graphs. Two things are worth noting.

Agentic AI reaches network biology. Ke Chen et al.'s Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI (the GenoMAS system) runs across roughly 1,300 disease-condition pairs. Together with the two LLM-agent papers in S2 and MedAgentBench v2 in S4, this makes four agentic papers across three sessions — the clearest signal at PSB 2026 that agents have moved from demo to method.

Provenance and interpretability for network algorithms. Tasnina et al.'s Provenance Tracing in Network Diffusion Algorithms asks which inputs actually drove a diffusion result — the network-biology version of the interpretability question the corpus keeps asking of foundation models.

The rest is solid applied work: Zhou et al.'s REPEL perturbs embeddings to improve protein-function learning; Krishnan et al. retrieve sparse perturbations for steady-state control; Orlenko et al. use a random-walk framework for Alzheimer's therapeutic targets; Rajagopalan et al.'s DRIVE-KG builds heterogeneous knowledge graphs over the Penn Medicine BioBank and Regeneron Genetics Center data for variant–phenotype discovery in understudied diseases.

Worth noting: organizer Joshua Welch is a single-cell integration methodologist (LIGER), but no single-cell paper landed in this session — see Themes for where the single-cell material actually went.

Papers

Title First author Pages PDF PMID
Provenance Tracing in Network Diffusion Algorithms Nure Tasnina 755–769 tasnina.pdf 41758183
REPEL — Random Embedding Perturbation for Enhanced Learning of Protein Function Di Zhou 770–783 zhou.pdf 41758184
Network Optimal Retrieval of Sparse Perturbations for Steady-State Control Krithika Krishnan 784–798 krishnan.pdf 41758185
Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI Ke Chen 799–814 chen_k.pdf 41758186
A Random-Walk-Based Learning Framework to Uncover Novel Gene Candidates for Alzheimer's Disease Therapy Alena Orlenko 815–829 orlenko.pdf 41758187
DRIVE-KG: Enhancing Variant-Phenotype Association Discovery in Understudied Complex Diseases Using Heterogeneous Knowledge Graphs Ananya Rajagopalan 830–848 rajagopalan.pdf 41758188

Connections to the corpus

Sources