Why ask whether existing drugs bind PACS1-R203W

The clinical observation that opens a question

Three observations motivate a drug-repurposing investigation specifically focused on the R203W mutation:

Empirical AED responses vary between PACS1 patients. Some respond well to phenytoin; some don't. Some respond to carbamazepine; many do not.
The mutation is recurrent in ~98% of patients. A single molecular target.
The mutation creates a new amino acid environment (Arg → Trp = a bulky, aromatic, hydrophobic residue) at a known protein-protein interaction site. Bulky aromatic side chains are known to create or modulate drug-binding pockets.

These three observations together suggest a clear hypothesis (H1): One or more FDA-approved drugs binds preferentially to PACS1-R203W vs wild-type PACS1, and could be selected (or de-selected) for clinical use based on this property.

What "preferential binding" would mean

In computational docking, preferential binding is quantified by Δ = ΔG(R203W) − ΔG(WT), where ΔG is the predicted binding free energy at a defined binding pocket. A negative Δ means the drug binds more tightly to R203W than to WT (mutation-selective). The reverse (Δ > 0) means it binds the wild-type better — not what we are looking for. A drug with Δ < −0.5 kcal/mol robustly (across multiple exhaustiveness levels and pocket conformations) is a candidate mutation-selective binder.

Falsifiable predictions

The hypothesis is testable both in silico (this project) and clinically (retrospective cohort analysis):

Setting	Prediction	How to test
In silico	A specific subset of FDA drugs has Δ < −0.5 robustly	virtual screening + robustness controls
In silico	The selectivity is reproducible at higher exhaustiveness	exhaust=16 and 32 docking
In silico	Random drugs do NOT show selectivity	random-drug control panel
In silico	Drugs from the same therapeutic class do NOT all bind	within-class controls
Clinical	PACS1 patients on the predicted-selective drug show better seizure outcomes	retrospective cohort analysis
Biochemical	Purified PACS1-R203W binds the drug with measurable affinity, WT does not	SPR or thermal shift assay

This work covers the in silico predictions and provides clinically testable hypotheses.

Why this is a useful path for ultra-rare diseases

For diseases with very small patient populations, drug discovery from scratch is economically infeasible. Wet-lab screening of 1,000+ compounds costs hundreds of thousands of euros and 1–2 years per protein. Computational repurposing can prioritize the 3–5 most promising existing drugs in days, for nearly zero cost. The downstream wet-lab and clinical validation then focuses on a tractable shortlist. The cost/time advantage is detailed on page 11.