# Bioinformatics-Driven Multi-Factorial Insight into α-Galactosidase Mutations

## TL;DR

This paper introduces a computational framework to better understand how mutations in the GLA gene affect alpha-galactosidase A, aiding in the treatment of Fabry disease.

## Contribution

A novel integrative framework combining structural and computational data to classify GLA gene mutations for precision medicine in Fabry disease.

## Key findings

- Amenable mutations preserve protein stability, while non-amenable ones cause structural destabilization.
- AlphaMissense predictions were compared with evolutionary and thermodynamic models to improve variant classification.
- Outlier variants with conflicting predictions were identified for further experimental validation.

## Abstract

Fabry disease is a rare genetic disorder caused by deficient activity of the lysosomal enzyme alpha-galactosidase A (AGAL), resulting in the accumulation of globotriaosylceramides (Gb3) in tissues and organs. This buildup leads to progressive, multi-systemic complications that severely impact quality of life and can be life-threatening. Interpreting the functional consequences of missense variants in the GLA gene remains a significant challenge, especially in rare diseases where experimental evidence is scarce. In this study, we present an integrative computational framework that combines structural, interaction, pathogenicity, and stability data from both in silico tools and experimental sources, enriched through expert curation and structural analysis. Given the clinical relevance of pharmacological chaperones in Fabry disease, we focus in particular on the structural characteristics of variants classified as “amenable” to such treatments. Our multidimensional analysis—using tools such as AlphaMissense, EVE, FoldX, and ChimeraX—identifies key molecular features that distinguish amenable from non-amenable variants. We find that amenable mutations tend to preserve protein stability, while non-amenable ones are associated with structural destabilisation. By comparing AlphaMissense with alternative predictors rooted in evolutionary (EVE) and thermodynamic (FoldX) models, we explore the relative contribution of different biological paradigms to variant classification. Additionally, the investigation of outlier variants—where AlphaMissense predictions diverge from clinical annotations—highlights candidates for further experimental validation. These findings demonstrate how combining structural bioinformatics with machine learning–based predictions can improve missense variant interpretation and support precision medicine in rare genetic disorders.