New Dashboard Brings Protein Structure Into the Hunt for Disease-Causing Gene Variants

For thousands of patients with rare immune disorders, a genetic test ends not with an answer but with a list of ambiguous results. Among the most frustrating entries on that list are variants of uncertain significance, or VUS: single-letter changes in a gene that sequencing machines detect effortlessly but that current interpretation frameworks cannot confidently label as harmless or harmful. A team of Brazilian researchers now reports a computational tool designed to shrink that gray zone by doing something most existing classifiers do not: looking at the three-dimensional structure of the protein the gene encodes.

The tool, called ProtSAVI, short for Protein Structure-Assisted Variant Interpretation, is described in BMC Bioinformatics by Nathália P.D. de Nigro, Paula F.C. Franklin, and colleagues working across Fundação Oswaldo Cruz, the Universidade Federal do Rio de Janeiro, and collaborating institutions in Rio de Janeiro. It is an R-based dashboard with a graphical user interface that assembles, in one place, the many strands of evidence clinicians and researchers typically have to gather manually: clinical genetics criteria, literature data, functional annotations, computational predictions, and structural analyses of the affected protein.

The problem ProtSAVI addresses is rooted in the way modern diagnostics work. Inborn errors of immunity are rare monogenic diseases in which a single altered gene disrupts immune system function. Next-generation sequencing has made it routine to sequence a patient’s genome and search for candidate mutations, but the interpretation bottleneck has not kept pace with the sequencing itself. When a laboratory finds a missense variant, a change that swaps one amino acid for another in the resulting protein, it must decide whether that swap explains the patient’s illness. Under the classification framework published by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology, known as ACMG/AMP, many such changes end up classified as VUS, a label that offers little guidance to physicians or families.

The prevalence of these uncertain calls is high enough that reclassification has become a discipline of its own. Every VUS represents a potential diagnosis left on the table, and every experimental test of a variant costs time, money, and laboratory capacity. Prioritizing which uncertain variants are most worth testing is therefore not an academic exercise; it directly determines how quickly patients receive answers. Existing automated classification tools largely follow the ACMG/AMP recommendations, but, as the authors note, they do not incorporate structural descriptors, the information about where a change sits in a folded protein and what that position implies about its function.

ProtSAVI was built to fill that gap. The dashboard integrates the ACMG/AMP evidence criteria with structural information to interpret and reclassify germline missense variants. Its analytical repertoire includes measures familiar to structural biologists: relative solvent accessibility, which describes how exposed a given amino acid is to the surrounding water; perturbation response scanning, a method for estimating how a mutation propagates structural effects through the protein; and assessments of whether a substitution might disrupt gain-of-function or loss-of-function mechanisms. By combining these biophysical descriptors with artificial intelligence approaches and functional data, the tool stratifies potentially pathogenic VUS, flagging the candidates most deserving of experimental validation.

The researchers evaluated their approach on a set of five genes associated with autoinflammatory inborn errors of immunity: ADA2, CDC42, STING1, TBK1, and WDR1. These genes encode proteins involved in immune signaling and inflammatory regulation, and mutations in them underlie a spectrum of rare autoinflammatory conditions. The team examined the variants in these genes that had been listed as VUS in ClinVar, the public archive of gene-disease relationships and variant interpretations, over the previous ten years. That search yielded 662 uncertain variants.

The results illustrate both the scale of the uncertainty problem and the tool’s ability to chip away at it. ProtSAVI provided usable interpretive information for 82 percent of those 662 cases, and within that subset it reclassified 37.75 percent of the variants. In practical terms, more than a third of the uncertain calls in these five immune genes could be moved toward a more definitive interpretation by combining structural analysis with existing evidence streams, without a single new experiment being run. For clinicians deciding which variants to send for functional assays, that kind of stratification can translate into substantially more efficient use of scarce validation resources.

What distinguishes ProtSAVI from a purely computational pipeline is its emphasis on transparency and usability. The tool is packaged as a dashboard rather than a command-line script, reflecting a deliberate design goal of bridging the gap between clinicians, who generate the variant lists, and structural biologists, who understand what the protein structures mean. It accepts standard variant data formats, including Variant Call Format files as well as comma-separated and tab-separated values, and it draws on resources such as the Protein Data Bank and literature-derived functional annotations. The authors position it as a framework in which the reasoning behind each reclassification is visible to the user, rather than delivered as an opaque score from a black-box model.

The work emerged from a broader national effort in Brazil. The research was carried out under the Rede Nacional de Genômica dos Erros Inatos da Imunidade, a national network for the genomics of inborn errors of immunity, with funding from agencies including FAPERJ, CNPq, the Brazilian health ministry’s science and technology department, and Fiocruz’s INOVA program. The team, co-directed by Roberta Soares Faccion, Zilton F.M. Vasconcelos, and Mauricio G.S. Costa, spans computational science, immunology, and clinical genetics, a mix that mirrors the interdisciplinary nature of the problem itself. The article is published open access, and the authors declare no competing interests.

The implications extend well beyond immune disorders. Because the dashboard is scalable and applicable to germline variants in other disease-related genes, the same structure-assisted logic could be applied wherever missense VUS accumulate, from cardiology to neurology to oncology genetics. The authors also outline a forward path: future versions of the tool will incorporate additional descriptors, including the effects of protein functional movements, capturing how a mutation perturbs not just a static structure but the dynamic motions a protein needs to do its job. As genome sequencing becomes ever more routine and the backlog of uncertain variants continues to grow, tools that can convert structural insight into clinical clarity are likely to become an increasingly standard part of the diagnostic toolkit, turning the ambiguous middle of a variant report into actionable answers for patients who have waited years for a diagnosis.

Subject of Research: Automated structure-assisted computational interpretation and prioritization of missense variants of uncertain significance in inborn errors of immunity

Article Title: Towards improved missense variant prioritization through automated structure-assisted interpretation

Article References: Towards improved missense variant prioritization through automated structure-assisted interpretation. (n.d.). https://doi.org/10.1186/s12859-026-06694-8

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06694-8

Keywords: missense variants, variants of uncertain significance, protein structure, inborn errors of immunity, ACMG/AMP, variant interpretation, ClinVar, autoinflammatory diseases, bioinformatics, structural biology, genetic diagnosis, variant prioritization