Research map / Image generation / Autoregressive and latent generation
MMCP-GEN: A Modality-Extensible Diffusion Language Model for Conditional Protein Sequence Generation
Zeyu An, Wanyu Lin, Feng Tan, Shujun Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| provide all conditions instead of only GO (ablation) | MMCP-GEN · GO-conditioned generation | MRR: 0.641 -> 0.873 | Improved |
| provide all conditions instead of only IPR (ablation) | MMCP-GEN · IPR-conditioned generation | macro F1: 0.861 -> 0.977 | Improved |
| provide all conditions instead of only EC (ablation) | MMCP-GEN · EC-conditioned generation | MRR: 0.789 -> 0.927 | Improved |
| add structure input with supervised fine-tuning of adapters and fusion (ablation) | MMCP-GEN · Inverse folding | AAR (%): 73.88 -> 77.49 | Improved |
| add all modalities beyond structure (ablation) | MMCP-GEN (SFT) · Inverse folding | AAR (%): 77.49 -> 78.17 | Improved |
| add joint generation-and-scoring (GS) loss (ablation) | MMCP-GEN (SFT, all conditions) · Inverse folding | AAR (%): 78.17 -> 78.66 | Improved |
| add GS loss on scTM and pLDDT (ablation) | MMCP-GEN (SFT, all conditions) · Inverse folding | scTM: 0.906 -> 0.912 | Improved |
| zero-shot use without fine-tuning adapters (ablation) | MMCP-GEN · Inverse folding | AAR (%): 73.88 vs CFP-GEN 77.01 | Mixed |
| MMCP-GEN versus CFP-GEN on GO-term MRR with only GO condition | MMCP-GEN (w/ GO) · GO-conditioned generation | MRR: 0.824 -> 0.641 | Hurt |
| MMCP-GEN versus CFP-GEN on IPR micro F1 with only IPR condition | MMCP-GEN (w/ IPR) · IPR-conditioned generation | micro F1: 0.961 -> 0.913 | Hurt |
Closest papers by meaning
- Diffusion on Language Model Encodings for Protein Sequence Generation (ICML 2025 · similarity 0.92)
- Diffusion Language Models Are Versatile Protein Learners (ICML 2024 · similarity 0.91)
- DPLM-2: A Multimodal Diffusion Protein Language Model (ICLR 2025 · similarity 0.91)
- CFP-Gen: Combinatorial Functional Protein Generation via Diffusion Language Models (ICML 2025 · similarity 0.91)
- Steering Generative Models with Experimental Data for Protein Fitness Optimization (NeurIPS 2025 · similarity 0.88)
- Controllable Protein Sequence Generation with LLM Preference Optimization (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
