1/ 🚀 Introducing AIDO.StructureDiffusion:.A generative model for structural protein design—enabling high-quality, controllable generation of monomers, complexes, and antibodies. 🧵

⏰ Happening tomorrow! Don’t miss the next Foundation Models for Biology (#FM4Bio) Seminar on Aug 5 at 9 AM PT. @abhinadduri will present “Predicting Cellular Responses to Perturbation Across Diverse Contexts With STATE.” STATE is a machine learning architecture for predicting

4/ This work strengthens AIDO’s vision: multiscale foundation models for predicting, simulating, and programming biology across contexts, from molecules to phenotypes. Read more ↓

3/ Why this matters for AIDO:. • Generating personalized models on-demand for every individual.• Unified ML paradigm for multi-modal data integration.• Information flow from molecular to cellular scale improves disease models. Paper 🔗

2/ Traditional methods miss patient-specific details. Contextualized models infer sample-specific GRNs, unlocking a more precise understanding of tumor biology.

1/ How can we model each tumor’s unique biology instead of relying on one-size-fits-all approaches?. In our latest blog post, we highlight findings from the recent PNAS paper with GenBio AI Research Scientist @probablybots and Co-Founder and Chief Scientist @ericxing, showing how

📡🧬 Join us for the next #FM4Bio Seminar on Aug 5 at 9 AM PT featuring @abhinadduri. Abhi will present “Predicting Cellular Responses to Perturbation Across Diverse Contexts With STATE.” This work introduces a machine learning architecture that predicts cellular responses to

5/ This research supports GenBio AI’s mission to develop foundation models that deliver precision, scale, and real-world impact in biology and medicine. Full paper📄:

4/ A multimodal AI foundation model trained with self-supervised learning on diet and glucose data outperformed existing methods in disease prediction. This brings us closer to digital twins for health.

3/ Key findings:. • Phenotype variation by age and ethnicity. • Biological age is not the same as chronological age. • Microbiome and disease associations. • Lifestyle and health correlations. • Individualized hypotheses for disease mechanisms.

2/ The HPP followed approximately 28,000 participants using:. • Continuous glucose and sleep monitoring. • Imaging and multi-omics. • Lifestyle and nutrition tracking. • Blood and immune profiling. A breakthrough dataset for understanding the health and disease continuum.

1/ 🚨 Just published in @NatureMedicine . An international team of researchers, including GenBio AI Co-Founder @ericxing, Senior Scientific Fellow @segal_eran, @MBZUAI, and others, shares a major advancement from the Human Phenotype Project (HPP). 📄 🎥

RT @EuxhenH: Excited to be at #ISMB2025 in Liverpool, UK this week! If you’re around, I’d love to connect and chat about foundation models,….

RT @cmuptx: If you are at ICML2025, welcome to attend our workshop "Multi-modal Foundation Models and Large Language Models for Life Scienc….

Thank you to everyone who joined us at GenBio AI’s private reception in Vancouver, BC. We’re grateful to be surrounded by such a vibrant, forward-looking scientific community in AI and biology. Stay tuned and learn more:

6/ This marks a step forward in fully integrated AI-guided protein design, where the design is no longer isolated but embedded within a broader biological context. Read the blog: .🔗

5/ Why it matters:. Controllability is key to molecular design. → You can guide the structure via the CAT hierarchy.→ Annotate CDRs, species, and chain types.→ Tune for loop length and developability.→ Build entire scFvs, not just fragments.

4/ It learns from real-world diversity. We trained the base model on AFDB and RCSB structures, weighted by structural motifs. Fine-tuned on SAbDab and interacting chains from RCSB, the model excels at producing viable backbones for nanobodies and fragments.

3/AIDO.StructureDiffusion enables end-to-end antibody generation, not just loop design or masked editing. You can design antibodies from scratch, guided by function, structure, and developability.

2/ What it does:. → Residue-wise CAT class conditioning.→ Precision CDR and light chain control.→ scFv generation.→ Multi-domain single-chain designs.→ Immunoglobulin-specific tuning.