Research map / Probabilistic and scientific modeling / Molecules and proteins
Interpolation-Based Conditioning of Flow Matching Models for Bioisosteric Ligand Design
Yael Ziv, Martin Buttenschoen, Lukas Scheibelberger, Brian Marsden, Charlotte Deane
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 |
|---|---|---|---|
| hard-replace coordinates (clean anchor) rather than noising coordinates, as in inpainting (ablation) | Replacement Guidance · DiffLinker test subset | % Valid: 35.6 -> 58.6 | Improved |
| use Replacement Guidance instead of fully noised inpainting (ablation) | Replacement Guidance · DiffLinker test subset | % Valid: 32.0 -> 58.6 | Improved |
| noise only bond logits instead of hard replacing them (ablation) | Replacement Guidance · DiffLinker test subset | % Valid: 58.6 -> 59.3 | No clear effect |
| add atom types to coordinate conditioning (ablation) | Replacement Guidance · DiffLinker test subset | % Valid: 86.3 -> 75.7 | Hurt |
| add bond connectivity to coords + atoms conditioning (ablation) | Replacement Guidance · DiffLinker test subset | SCRDKit: 0.43 -> 0.55 | Improved |
| raise t_free from 0.95 to 1.00 (hold fragment constraints longer) (ablation) | Replacement Guidance · DiffLinker test subset | % Valid: 72.6 -> 58.6 | Hurt |
| raise t_free from 0.85 to 1.00 (ablation) | Replacement Guidance · DiffLinker test subset | SCRDKit: 0.49 -> 0.55 | Improved |
| lower restart time tau (smaller s index set to 0.70 vs 0.80) in Interpolate-Integrate (ablation) | Interpolate-Integrate · GEOM-Drugs seeds | Valid & Unique & Novel (%): 20.36 -> 84.24 | Improved |
| increase seed noise sigma from 0.00 to 0.25 (ablation) | Interpolate-Integrate · GEOM-Drugs seeds (s=0.72) | % Valid: 87.83 -> 80.18 | Hurt |
| increase seed noise sigma from 0.00 to 0.25 (ablation) | Interpolate-Integrate · GEOM-Drugs seeds (s=0.72) | Valid & Unique & Novel (%): 30.52 -> 69.26 | Improved |
| increase restart step s from 68 to 80 (more noise) for disconnected seeds (ablation) | Interpolate-Integrate · DiffLinker 3-fragment seeds | % Valid: 58.1 -> 82.2 | Improved |
| increase number of seed fragments from 2 to 10 (ablation) | Interpolate-Integrate · Mpro 25-35 atoms | % Valid: 84.0 -> 65.0 | Hurt |
Closest papers by meaning
- Unified Guidance for Geometry-Conditioned Molecular Generation (NeurIPS 2024 · similarity 0.89)
- ShEPhERD: Diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design (ICLR 2025 · similarity 0.89)
- Harmonic Self-Conditioned Flow Matching for joint Multi-Ligand Docking and Binding Site Design (ICML 2024 · similarity 0.88)
- Multi-domain Distribution Learning for De Novo Drug Design (ICLR 2025 · similarity 0.88)
- Controllable 3D Molecular Generation for Structure-Based Drug Design Through Bayesian Flow Networks and Gradient Integration (NeurIPS 2025 · similarity 0.88)
- Compositional Flows for 3D Molecule and Synthesis Pathway Co-design (ICML 2025 · similarity 0.87)
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.
