STRC Research

MRGPRX2 AF-M screen

2 min read

MRGPRX2 AF-M screen

What it does

R scripts for parsing AlphaFold-Multimer / ColabFold multimer outputs from nanobody-target runs and calculating interface features used in Harvey et al. 2026: pTM, interface PAE, interface pLDDT, pDockQ, model support, and the composite LCF score.

Source

Install

git clone https://github.com/kruselab/MRGPRX2-AF-M-screen.git

R dependencies listed by the upstream README: tidyverse, cowplot, RColorBrewer, jsonlite, micropan, Biostrings, GGally, plotROC, here.

Commands

The upstream README exposes the scoring entry point:

source("scoring.R")
run_pipeline("data/")

Expected outputs:

OutputMeaning
res_pair_data/*.csvPer-model interchain residue-pair information
score_data/complex_scores.csvOne row per AF-M run with complex-level scores
score_data/model_scores.csvOne row per AF-M model with model-level scores

STRC use

Use as a method reference, not an off-the-shelf therapeutic engine.

  • h09 optional Phase 4o/4p: parse AF3/AF-M interface confidence features for existing tail91 x TMEM145 / actin jobs and matched negative controls.
  • h26 future-only: if a homodimer candidate ever passes G1/G2, interface feature extraction can supplement contact counts.

Do not use for

  • Absolute Kd prediction.
  • Promoting STRC/TMEM145 binders from AF-M scores alone.
  • Claiming nanobody discovery is validated for STRC, TMEM145, actin, RADA16, or other non-GPCR targets.

The paper’s own benchmark reports strong GPCR top-5% enrichment but poor soluble/non-GPCR transfer.

Connections

Graph View

Backlinks