Model Requests

Antibody-focused developability model with partial solubility/HIC-RT/viscosity signal; very slow throughput. Primary use cases: Antibody candidate developability triage HIC-RT / viscosity risk cross-check Shortlist “last mile” evaluation

Strong for solubility and expression, especially on enzymes/general proteins; slow throughput. Primary use cases: Joint solubility + expression feasibility scoring Enzyme programs where expression yield matters Secondary/confirmatory scoring after fast screens

Strong Tm/thermostability predictor; generalizable to antibodies, enzymes, and general proteins; fast throughput. Primary use cases: Predicting thermostability/Tm shifts for mutants Stability screening in antibody & enzyme programs Benchmarking against BioLM stability models

Solubility predictor with strong expression/enzyme signal; generally applicable; fast throughput. Primary use cases: Enzyme solubility + expression feasibility screening Complementary solubility baseline for non-antibody proteins Multi-objective developability alongside BioLM solubility/stability models

CamSol is a widely used solubility predictor. In our summary matrix it’s strong for Solubility and works well for Antibodies and General proteins, with fast throughput. It’s a good baseline/industry reference for solubility risk flagging early in design and developability. Primary use cases: Early developability screening for soluble expression Antibody and general protein solubility triage Baseline comparison against BioLM native solubility models

BoltzGen is a new generative model for designing protein and peptides of any modality to bind a wide range of biomolecular targets. https://boltz.bio/boltzgen

Integrate DiffDock-L, a new version of DiffDock, for High-Accuracy Diffusion-Based Ligand Docking . It provides significant performance improvement and generalization capacity over previous state-of-the-art for small-molecule docking. Paper/GitHub: gcorso/DiffDock .

Diffusion Sequence Models for Enhanced Protein Representation and Generation. https://arxiv.org/abs/2506.08293 @misc{hallee2025diffusionsequencemodelsenhanced, title={Diffusion Sequence Models for Enhanced Protein Representation and Generation}, author={Logan Hallee and Nikolaos Rafailidis and David B. Bichara and Jason P. Gleghorn}, year={2025}, eprint={2506.08293}, archivePrefix={arXiv}, primaryClass={q-bio.BM}, url={https://arxiv.org/abs/2506.08293}, }

Add LightDock for Flexible Protein-Protein and Protein-Ligand Docking . A multi-scale approach based on the GSO algorithm, known for its speed, versatility, and ability to handle flexible docking scenarios. Paper/GitHub: lightdock/lightdock .

Add Transpolymer to enable Polymer Science and Biomaterials Modeling . This model extends BioLM's utility by allowing users to model and predict properties of non-biological polymer sequences. Paper/GitHub: Pratham-taneja/TransBERT-Polymer-property-prediction-Research-paper .