AlgorithmicRx: AI-Designed Molecular Rescue for Duchenne Muscular Dystrophy

We're building a computational proof-of-concept (POC) that uses AlphaFold 3 to discover compensatory proteins capable of restoring a critical structural link broken in DMD: from disease hypothesis to in silico validation

The Problem

Duchenne Muscular Dystrophy (DMD) is caused by mutations in the dystrophin gene , that leave the protein truncated or missing entirely. Dystrophin normally acts as a molecular shock absorber, anchoring the cell's internal actin skeleton to the Dystrophin-Associated Protein Complex (DAPC) spanning the muscle cell membrane. When that link breaks, β-dystroglycan is left unanchored, and the membrane loses its structural integrity under the mechanical stress of muscle contraction — driving the progressive muscle damage seen in DMD.

Our Approach

AlgorithmicRx is an early-stage computational pipeline built to identify compensatory rescue candidates: small proteins or peptides designed to bind the exposed pocket on β-dystroglycan and computationally restore the lost mechanical linkage, modeled as an alternative route around the broken dystrophin connection.

The pipeline currently:

● Models the healthy (wild-type) DAPC assembly — mapping how dystrophin, β-dystroglycan, the sarcoglycan complex, and syntrophin fit together in the intact system.

● Models the disease state, simulating the structural consequences of dystrophin truncation on the complex.

● Generates candidate rescue binders through motif-grafted peptide library design targeting the exposed β-dystroglycan interface.

● Screens candidates in silico using AlphaFold 3, evaluating predicted binding confidence (ipTM) and interface reliability (pLDDT) for each design.

● Analyzes interface quality — inter-chain contact hotspots, predicted alignment error (PAE), and per-residue confidence — to rank the most promising leads.

Where We Are Today (MVP Stage)

This is a proof-of-concept, in silico research pipeline — not a validated therapeutic. Our early results show promising computational binding signals (e.g., ipTM ≈ 0.85 with high-confidence interface predictions) for select rescue candidates against β-dystroglycan. The next steps are expanding candidate libraries, tightening validation criteria, and eventually pursuing experimental (wet-lab) validation of top candidates.

We're sharing this as we build toward the next stage which is running this POC in silico research with other open source protein prediction models like ESMFold2 , OpenFold3 or Boltz-2 / Boltz-1

The Pipeline

Component	What it does
fetch_dapc.py	Pulls UniProt sequences and structural data for all DAPC components
af3_json_generator.py	Builds AlphaFold 3 input configurations for structure prediction
binder_designer.py	Generates and ranks candidate rescue-peptide libraries
analyze_ppi.py	Scores predicted interfaces using pLDDT, PAE, and contact-hotspot analysis
dapc_colab_pipeline.py	Notebook-based workflow modeling disease-state mutations end-to-end

Why It Matters

Gene therapy has already made real progress here: Elevidys (delandistrogene moxeparvovec), FDA-approved for DMD, delivers a shortened, functional "micro-dystrophin" gene via an AAV viral vector to help muscle cells produce a functional protein and slow degeneration. But real-world adoption still faces meaningful barriers — high cost, eligibility limits for non-ambulatory or older patients, and immune responses against the viral vector that can exclude patients from treatment or prevent re-dosing. AlgorithmicRx's compensatory-binder approach is exploring a different route around some of these constraints — not a replacement for gene therapy, but a complementary hypothesis worth testing computationally, faster and cheaper than jumping straight to the lab.

Get Involved

We're at the MVP stage and looking to connect with:

Structural biologists and computational chemists interested in DAPC/DMD research

Wet-lab collaborators who could validate top candidates experimentally

Anyone building in the rare-disease / computational drug discovery space

Request the Technical Writeup — for structural biologists, computational chemists, and wet-lab collaborators who want to evaluate or extend the work

Follow Our Progress — for anyone who wants updates

Disclaimer

This is an early-stage research project under AlgorithmicRx's DMD pipeline All results described are computational/in silico predictions and have not been experimentally or clinically validated. This is not a medical treatment and should not be interpreted as therapeutic advice.