Background/Objectives: Congenital Myasthenic Syndromes (CMS) are caused by variants in >30 genes with increasing numbers of variants of unknown significance (VUS) discovered by next generation sequencing. Establishing VUS pathogenicity requires in vitro studies that slow diagno-sis and treatment initiation. Recent protein structure prediction softwares, AlphaFold2/ColabFold, have revolutionised structural biology; such predictions have also been leveraged in AlphaM-issense which predicts ClinVar variant pathogenicity with 90% accuracy. Few reports, however, have tested these tools on rigorously-characterised clinical data. We therefore assessed ColabFold and AlphaMissense as diagnostic aids for CMS, using variants of the CHRN genes that encode the nicotinic acetylcholine receptor (nAChR).
Methods: Utilising a dataset of 61 clinically-validated CHRN variants, (1) we evaluated the pos-sibility of a ColabFold metric (either predicted structural disruption, prediction confidence, and prediction quality), that distinguishes variant pathogenicity. (2) We assessed AlphaMissenses’ ability to differentiate variant pathogenicity. (3) We compared AlphaMissense to existing patho-genicity prediction programmes, AlamutVP and EVE.
Results: Analysing variant effects on ColabFold CHRN structure prediction, prediction confi-dence, and prediction quality, doesn’t yield any pathogenicity-indicative metric. However, Al-phaMissense predicts variant pathogenicity with 63.93% accuracy on our dataset – a much greater proportion than AlamutVP (27.87%) and EVE (28.33%).
Conclusions: Emerging in silico tools can revolutionise genetic disease diagnosis – however, im-provement, refinement, and clinical validation is imperative prior to practical acquisition.
