BioMap Research Wins First Place at Arc’s Virtual Cell Challenge 2025
We are proud to announce that BioMap Research took first place at Arc’s Virtual Cell Challenge 2025 with our independently developed model, xTrimoSCPerturb, outperforming over 1,200 teams from 114 countries.
What once required years of trial-and-error can now be designed, evaluated, and optimized computationally—turning target discovery into an iterative, scalable engineering problem.
De novo antibody design is only as good as its validation — especially when the target matters therapeutically.
LGR5 is a critical stem cell marker driving tumor growth in colorectal, gastric, and ovarian cancers.
The result: epitope-specific nanobodies with nanomolar affinity and Tm >65° C - designed computationally, validated structurally, ready for therapeutic development.
We also develop a novel in-context learning (ICL) paradigm for protein optimization using its reasoning. It offers a scalable framework for protein modeling/optimization. Paper link:https://t.co/UeLqHzJ0xC
We present ProteinReasoner, a multi-modal protein language model using chain-of-thought (CoT) with evolutionary profiles to link structure/sequence, boosts zero-shot performance in structure/folding/fitness prediction tasks, beating models like ESM3/DPLM-2.
🧬 NEW RESEARCH: We systematically benchmarked 4 leading inverse folding models for antibody CDR sequence design—critical for advancing antibody engineering and the de novo design of therapeutic mAbs.
🔬 Read the full study here: https://t.co/RJEfCMzaqA