Hit-to-lead progression on Hit-to-lead 30 protein targets (test)
100ValidityMolLingo
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MolLingoBase LLM=GPT-5.4, Input Context=docking-guided biological context, Architecture=Agentic Multi-Agent2026.05 | 100 | 10.441 | 65.19 | 0.63 | 3.343 | |
| Claude-4.6-SonnetInput Context=raw SMILES + protein name2026.05 | 100 | 4.308 | 16.527 | 0.864 | 2.846 | |
| Qwen2-14B-InstructInput Context=raw SMILES + protein name2026.05 | 100 | 4.874 | 20.471 | 0.774 | 2.96 | |
| Qwen2-7B-InstructInput Context=raw SMILES + protein name2026.05 | 100 | 4.353 | 20.01 | 0.758 | 2.935 | |
| GPT-5.4Input Context=raw SMILES + protein name2026.05 | 98.67 | 2.437 | 75.871 | 0.769 | 2.402 | |
| GPT-4.1Input Context=raw SMILES + protein name2026.05 | 97.33 | 2.327 | 73.629 | 0.78 | 2.867 | |
| Gemini-3-ProInput Context=raw SMILES + protein name2026.05 | 94 | -1.361 | 14.103 | 0.878 | 3.104 |