AI Paper Atlas

Research map / Large language models / Agents, code and math

PDAgent: An LLM-Driven Autonomous Agent Framework Towards *In Silico* Protein Design via Directed Mutation

Song Ouyang, Zhijie Dong, Yong Luo, Kehua Su, Huangxuan Zhao, Miaojing Shi, Bo Du

ICML 2026 · Large language models › Agents, code and math · approach: LLM-guided program synthesis

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 changedWhereMeasuredEffect
add the iteration loop (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 81.85 -> 90.60Improved
add history memory (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 89.36 -> 90.60Improved
separate THINK and PLAN phases (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 86.53 -> 90.60Improved
add conservation guard (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 89.65 -> 90.60Improved
add conservation guard (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3foldability: 88.33 -> 87.69Hurt
add mutation loop to template-only output (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 82.24 -> 90.60Improved
use knowledge-guided mutation instead of random generation (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3average CSR: 55.46 -> 90.60Improved
use knowledge-guided mutation instead of random generation (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3foldability: 30.24 -> 87.69Improved
increase candidate templates from 1 to 3 (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasksCSR: 72.5 -> 90.6Improved
increase candidate templates from 1 to 3 (ablation)PDAgent (Qwen-3-30B) · 100 design tasksCSR: 67.3 -> 78.9Improved
set conservation threshold to 0.6 rather than 0.2 or 0.8 (ablation)PDAgent (LLaMA-3.1-8B) · 100 design tasks, n=3CSR and pLDDTImproved
replace best direct-LLM baseline Gemini-3-Pro with PDAgentPDAgent (DeepSeek-V3.2) · 100 design tasksaverage CSR: 82.55 -> 91.86Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy