Release endpoints
Support fixed, left-released, right-released, and both-released modes by promoting terminal anchor pairs to SE(3) rigid bodies.
All projects
Molecular modeling
Inverse-kinematics protein backbone sampling
A geometric sampler for loops, protein fragments, and intrinsically disordered regions
4
endpoint modes
30 aa
loop sampling
56 aa
IDR sampling
Context
The problem
Closed-loop samplers preserve both endpoints, but many conformational-search problems require one or both ends of a backbone segment to move. Releasing these constraints expands the search space while retaining the local geometry and tripeptide-reconstruction machinery of loop closure.
The project treats released terminal anchors as free rigid bodies, preserves feasible tripeptide reconstruction along candidate motions, and uses multiple shifted frames so every backbone atom experiences both rigid-body and closure moves.
Method
How the system works
Propose feasible motion
Generate Hit-and-Run trajectories for peptide bodies and released anchors, then intersect them with geometric feasibility domains.
Resample all atoms
Cycle through three shifted tripeptide frames so the full backbone is updated rather than leaving a persistent subset rigid.
Filter and validate
Use spatial-hash clash detection, side-chain completion, restrained minimization, torsional validation, clustering, and representative selection.
Role
My contribution
- Optimized the C++ protein loop sampler and extended it to support fixed or independently released termini.
- Implemented released-anchor SE(3) proposals, boundary reconditioning, multi-frame resampling, and geometric-hash collision checks.
- Built the full-atom validation stack using AttnPacker, OpenMM, PhiSiCal-Checkup, MDS, and medoid-aware clustering.
- Designed controlled comparisons against BBFlow for short loops and idpGAN-derived ensembles for an intrinsically disordered region.
Result
Outcomes and scientific value
✓
Provides a unified geometric proposal framework spanning closed loops, free backbone fragments, and longer disordered segments.
✓
Combines permissive conformational exploration with explicit downstream evaluation of sterics, local energy, torsional quality, and ensemble diversity.
✓
Current benchmarking covers a 30-residue loop and a 56-residue androgen-receptor IDR; the manuscript remains under development.