Research map / Reinforcement learning / Visual and representation RL
Interpretability Driven Evolutionary Approach for the Design of Biological Sequences
Akash Pandey, Wei Chen, Sinan Keten
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 |
|---|---|---|---|
| add exploratory mutations to IDEAS (IDEAS vs IDEAS-X) (ablation) | IDEAS · AUC across 8 datasets | AUC: mean AUC differs by less than 2% | No clear effect |
| add exploratory mutations to IDEAS (ablation) | IDEAS · AUC-Diversity, B=50 | diversity: 41% higher diversity | Improved |
| select mutation positions by low attribution (tau1=1) instead of random (tau1=1000) (ablation) | IDEAS · GRAVY index | AUC: 32.60 -> 35.40 | Improved |
| select positions with low rather than high attribution (tau1=1 vs -1) (ablation) | IDEAS · GRAVY index | AUC: 19.74 -> 35.40 | Improved |
| select replacement motifs by high attribution (tau2=1) instead of random (tau2=1000) (ablation) | IDEAS · Binding affinity TF1 | AUC: 5.22 -> 5.80 | Improved |
| select high-attribution replacement motifs (tau2=1 vs -1) (ablation) | IDEAS · GRAVY index | AUC: 28.03 -> 35.40 | Improved |
| raise temperatures tau1,tau2 (more exploration) (ablation) | IDEAS · GRAVY index | AUC vs diversity | Mixed |
| increase number of exploitative mutations mu (ablation) | IDEAS · GRAVY index | AUC vs diversity | Mixed |
| swap COLOR for DeepLift, DeepLiftShap, GradientSHAP or Integrated Gradients (ablation) | IDEAS · GRAVY and Aliphatic index | AUC: average AUC differs by less than 2% | No clear effect |
| add label noise to XAI training data (eta 0 to 10) (ablation) | IDEAS · GRAVY index | AUC: 37.10 -> 34.22 | Hurt |
| extreme label noise (eta=50) making attributions random (ablation) | IDEAS · GRAVY index | AUC: 37.10 -> 26.44 | Hurt |
| increase motif size m from 1 to 2 (ablation) | IDEAS · Aliphatic index | AUC: AUC +23%, Diversity +36%; plateaus beyond | Improved |
Closest papers by meaning
- Designing Biological Sequences without Prior Knowledge Using Evolutionary Reinforcement Learning (AAAI 2024 · similarity 0.90)
- Tree Search-Based Evolutionary Bandits for Protein Sequence Optimization (AAAI 2024 · similarity 0.87)
- Knowledge-aware Reinforced Language Models for Protein Directed Evolution (ICML 2024 · similarity 0.87)
- Steering Generative Models with Experimental Data for Protein Fitness Optimization (NeurIPS 2025 · similarity 0.86)
- Improved Off-policy Reinforcement Learning in Biological Sequence Design (ICML 2025 · similarity 0.86)
- ProSpero: Active Learning for Robust Protein Design Beyond Wild-Type Neighborhoods (NeurIPS 2025 · similarity 0.86)
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.
