ISMB/BOSC 2026
📝 Event Description
ISMB/ECCB 2026 (Intelligent Systems for Molecular Biology / European Conference on Computational Biology) is the world's largest bioinformatics conference, bringing together researchers across computational biology, bioinformatics, and open science. Among the many ISMB tracks, BOSC (Bioinformatics Open Source Conference) is dedicated to open-source software and open science in bioinformatics. This year, there will also be an NIH/ODSS track titled "Social and Technical Infrastructure to Advance Health Research" that will explore the platforms, standards, policies, and communities that enable modern biomedical research.
More information is available on the ISMB 2026 website, the BOSC 2026 website, and the NIH/ODSS program page.
💡 Our Role / Contribution
Our team will have a strong presence with four talks, four posters, and a CoFest project across the BOSC and NIH/ODSS tracks.
🎤 Talks & Poster — Bhavesh Patel
Posters.science: Building Open Infrastructure for FAIR and AI-Ready Scientific Posters
In the BOSC track, Bhavesh will give a short talk and present a poster on Posters.science, a free, open-source platform designed to make scientific posters FAIR and AI-ready. Posters.science allows researchers to upload a poster PDF, review auto-extracted structured metadata via an LLM-based pipeline, and publish to an open repository in just a few steps.
- Project website: posters.science
Clinical Dataset Structure: A Universal Standard for Structuring Clinical Research Datasets
In the NIH track, Bhavesh will give a talk on the Clinical Dataset Structure (CDS), a standardized framework for organizing multimodal clinical research data such that they are FAIR and AI-ready, which is developed as part of the AI-READI project.
- Project website: cds-specification.readthedocs.io
🎤 Short Talk & Poster — Jamey O'Neill
PosterSentry: A Lightweight Multimodal Classification System for Scientific Poster Quality Assurance in Open Repositories
Jamey will present PosterSentry, a lightweight, GPU-free classification tool that determines whether a given PDF is a valid scientific poster. Built to address widespread mislabeling in open repositories, PosterSentry combines text, visual, and structural features to accurately classify PDFs at scale. The model is fully open-source (MIT) and published on HuggingFace. It serves as the quality-assurance layer at the front of the Posters.science pipeline.
- Project website: huggingface.co/fairdataihub/poster-sentry
- Source code: github.com/fairdataihub/poster-sentry-training
🎤 Short Talk & Poster — Nahid Zeinali
Evaluating the Performance of Large Language Models in Drafting Data Management Plans
Nahid will present a study evaluating how well LLMs can draft NIH-compliant Data Management Plans (DMPs) off-the-shelf. The study compared GPT-4.1 and Llama 3.3 against human-written DMP examples from previously funded NIH proposals, using both automatic metrics and expert human review. The results suggest LLMs can accelerate the creation of initial drafts, but expert review remains essential before submission.
- Source code: github.com/fairdataihub/nih-dmp-llm-generation
- Dataset: doi.org/10.5281/zenodo.19456208
🪧 Poster — Aydan Gasimova
Clinical Dataset Structure: A Universal Framework for Structuring Clinical Research Datasets
Aydan will present the Clinical Dataset Structure (CDS), an open-source (CC-BY-4.0) standardized framework for organizing multimodal clinical research data and metadata. Developed through the AI-READI project, the CDS unifies existing domain-specific standards (e.g., BIDS), metadata schemas (DataCite, ClinicalTrials.gov), and AI/ML-focused standards (Datasheet, Healthsheet). The CDS specification is documented at cds-specification.readthedocs.io.
🤝 CoFest — Rescuing Scientific Posters from Obscurity
The entire team will participate in the BOSC CoFest (Collaborative Sprinting event), where we are proposing a sprint around Posters.science and related tools. Proposed directions include:
- Biomedical schema extensions: Enriching the poster JSON schema with bioinformatics-aware fields (linked datasets, tools/software, organisms, EDAM ontology terms, etc.)
- Benchmarking poster2json: Building a curated benchmark of annotated posters (e.g., from past BOSC conferences) to systematically evaluate and improve the LLM-based poster metadata extraction pipeline
- Smart semantic search: Prototyping a natural-language search feature over the 24k+ posters in the Posters.science registry
All participants are welcome. We are open to pursuing whatever directions the CoFest community finds most exciting!
📍 Location
Washington, DC
ISMB 2026 Website | BOSC 2026 Website
🗓 Date and Time
ISMB/ECCB 2026 runs July 12–16, 2026, in Washington, DC. The NIH/ODSS track in one day on July 13. BOSC is a multi-day COSI within ISMB (July 14-15). The CoFest (collaborative sprinting event) follows on July 17–18.
NIH track
- Monday July 13, 2026, 2:35–2:45 pm: Bhavesh's talk on CDS
- Monday July 13, 2026, 3:25–4:00 pm: Panel discussion
BOSC track
Talks:
- Wednesday July 15, 2026, 2:40-2:45 pm: Bhavesh's short talk on Posters.science
- Wednesday July 15, 2026, 3:20-3:25 pm: Nahid's short talk on using LLMs for drafting DMPs
- Wednesday July 15, 2026, 3:30-3:35 pm: James' short talk on PosterSentry
Posters:
- Tuesday July 14, 2026, at 10:00-11:00 am and 4:00-4:40 pm: Aydans' poster on CDS
- Wednesday July 15, 2026, at 10:00-11:00 am and 4:00-4:40 pm: James' poster on PostersSentry
- Thursday July 16, 2026, at 10:00-11:00 am and 4:00-4:40 pm: Bhavesh's poster on Posters.science
- Thursday July 16, 2026, at 10:00-11:00 am and 4:00-4:40 pm: Nahid's poster on using LLMs for drafting DMPs
🔗 Link to Materials
- CDS (Bhavesh): https://github.com/fairdataihub/CDS-ISMB-NIH-2026
- Posters.science (Bhavesh): https://github.com/fairdataihub/posters-science-ISMB-BOSC-2026