Characterisation of patients with Alpha-1 antitrypsin deficiency using unsupervised machine learning tools

Area of Science:

• Medical Informatics
• Genetics
• Pulmonology

Background:

• Alpha-1 antitrypsin deficiency (AATD) is an underdiagnosed genetic disorder causing emphysema and liver disease.
• Clinical heterogeneity and individual variability complicate AATD risk stratification and personalized treatment.
• Artificial Intelligence (AI) and Machine Learning (ML) offer potential for novel AATD clinical characterization.

Purpose of the Study:

• To explore the feasibility of applying ML for clinical characterization of AATD patients.
• To segment AATD patients into clinically meaningful clusters using unsupervised learning.
• To assess the potential of ML-driven patient stratification for personalized medicine in AATD.

Main Methods:

• A pilot, single-center, cross-sectional study included 210 AATD patients from the EARCO registry.
• The unsupervised k-prototypes algorithm was used to identify clusters based on clinical and demographic data.
• Statistical analyses (Mann-Whitney U, Chi-square) compared functional and clinical differences among identified clusters.

Main Results:

• Five distinct patient clusters were identified based on clinical and genetic profiles.
• Clusters included patients with emphysema, bronchiectasis, asymptomatic genotypes, COPD with comorbidities, and altered hepatic profiles.
• Significant differences in pulmonary function, AAT levels, and comorbidities were observed across clusters.

Conclusions:

• ML can effectively segment AATD patients into clinically meaningful subgroups.
• This AI-driven approach holds promise for guiding therapeutic decisions and improving patient follow-up.
• ML-based stratification can support the design of targeted, cluster-based multicenter trials for AATD.