Papers › Knowledge-Design: Pushing the Limit of Protein Design via Knowledge Refinement

Knowledge-Design: Pushing the Limit of Protein Design via Knowledge Refinement

20 May 2023arXiv:2305.15151archive 2025-07-28

Zhangyang Gao, Cheng Tan, Stan Z. Li

Recent studies have shown competitive performance in protein design that aims to find the amino acid sequence folding into the desired structure. However, most of them disregard the importance of predictive confidence, fail to cover the vast protein space, and do not incorporate common protein knowledge. After witnessing the great success of pretrained models on diverse protein-related tasks and the fact that recovery is highly correlated with confidence, we wonder whether this knowledge can push the limits of protein design further. As a solution, we propose a knowledge-aware module that refines low-quality residues. We also introduce a memory-retrieval mechanism to save more than 50\% of the training time. We extensively evaluate our proposed method on the CATH, TS50, and TS500 datasets and our results show that our Knowledge-Design method outperforms the previous PiFold method by approximately 9\% on the CATH dataset. Specifically, Knowledge-Design is the first method that achieves 60+\% recovery on CATH, TS50 and TS500 benchmarks. We also provide additional analysis to demonstrate the effectiveness of our proposed method. The code will be publicly available.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

A4Bio/OpenCPD officialmentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Protein DesignRetrievalWord Sense Disambiguation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Protein Design CATH 4.2 Knowledge-Design Perplexity 3.46 #1 of 8 Archive leaderboard report
Protein Design CATH 4.2 Knowledge-Design Sequence Recovery %(All) 60.77 #1 of 8 Archive leaderboard report
Protein Design CATH 4.2 PiFold Perplexity 4.55 #2 of 8 Archive leaderboard report
Protein Design CATH 4.2 PiFold Sequence Recovery %(All) 51.66 #2 of 8 Archive leaderboard report
Protein Design CATH 4.2 ProteinMPNN Perplexity 4.61 #3 of 8 Archive leaderboard report
Protein Design CATH 4.2 ProteinMPNN Sequence Recovery %(All) 45.96 #3 of 8 Archive leaderboard report
Protein Design CATH 4.2 GVP Perplexity 5.36 #4 of 8 Archive leaderboard report
Protein Design CATH 4.2 GVP Sequence Recovery %(All) 39.47 #4 of 8 Archive leaderboard report
Protein Design CATH 4.2 GCA Perplexity 6.05 #5 of 8 Archive leaderboard report
Protein Design CATH 4.2 GCA Sequence Recovery %(All) 37.64 #5 of 8 Archive leaderboard report
Protein Design CATH 4.2 AlphaDesign Perplexity 6.3 #6 of 8 Archive leaderboard report
Protein Design CATH 4.2 AlphaDesign Sequence Recovery %(All) 41.31 #6 of 8 Archive leaderboard report
Protein Design CATH 4.2 StructGNN Perplexity 6.4 #7 of 8 Archive leaderboard report
Protein Design CATH 4.2 StructGNN Sequence Recovery %(All) 35.91 #7 of 8 Archive leaderboard report
Protein Design CATH 4.2 GraphTrans Perplexity 6.63 #8 of 8 Archive leaderboard report
Protein Design CATH 4.2 GraphTrans Sequence Recovery %(All) 35.82 #8 of 8 Archive leaderboard report
Protein Design CATH 4.3 GVP-large Perplexity 6.17 #1 of 2 Archive leaderboard report
Protein Design CATH 4.3 GVP-large Sequence Recovery %(All) 39.2 #1 of 2 Archive leaderboard report
Protein Design CATH 4.3 ESM-IF Perplexity 6.44 #2 of 2 Archive leaderboard report
Protein Design CATH 4.3 ESM-IF Sequence Recovery %(All) 38.3 #2 of 2 Archive leaderboard report
Word Sense Disambiguation TS50 SPIN Sequence Recovery %(All) 30.3 #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

fail

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections