Evaluating Potential Subpopulation of ABCA4 Pathogenic Variants Based on Protein Structural Distribution and Patient Phenotypes

• Michael Sturtevant, Biological Sciences, University of Delaware
• Senem Cevik, Biological Sciences, University of Delaware

• Esther Biswas-Fiss, , University of Delaware

Evaluating Potential Subpopulation of ABCA4 Pathogenic Variants Based on  Protein Structural Distribution and Patient Phenotypes

Michael Sturtevant, Senem Cevik and Esther Biswas-Fiss

Introduction: The ATP-binding cassette subfamily A member 4 (ABCA4) gene encods a transmembrane transporter that removes toxic byproducts in photoreceptor cells to maintain vision. A major challenge lies in comprehensively understanding how variations in the ABCA4 gene lead to disease. Among the ABCA4 variants, a significant portion (n=1653) are missense mutations, where one amino acid is replaced by another. Nearly half of these are classified as variants of uncertain significance (VUS) variants, with an additional 12% presenting conflicting interpretations (CI).

ClinVar is a repository that categorizes human genetic variations based on their clinical significance: pathogenic, likely pathogenic, uncertain significance, likely benign, and benign. This study investigates whether pathogenic variants of known significance (VKS) form subpopulations based on their impacts on clinical phenotypic characteristics.  

Approach: We focused on missense ABCA4 variants listed as pathogenic in the ClinVar database. To identify potential subpopulations of these variants based on their impacts on the patient phenotypes, we investigated metrics in the literature indicating disease severity, such as age of onset and clinical presentation, considering the entire ABCA4 genetic profile of those patients. Information on VKS will prove valuable in creating machine learning-based algorithms to assess ABCA4 VUS.

Results: The collected data reveals that the pathogenic/likely pathogenic mutations are evenly distributed throughout the protein, with patient data showing patterns in which variants associate with certain stages of macular degeneration.

Conclusion: This observation supports the hypothesis that training a computer program to identify pathogenic variants after being fed a list of VKS would not only be more efficient at identifying patterns in pathogenic mutations, but required. The pathogenic VKS are evenly spread out among the protein, making manual analysis of VUS infeasible.