A Framework for Variant Filtering to Improve Diagnostic Outcomes in Genome Sequencing

Multi-step pre-annotation filtering (conservation, allele frequency, splicing) followed by phenotype-based prioritization, targeting non-coding regions

The authors propose SMART filtering (Systematic, Multi-step, Automated, Rare variant prioritization Tool), a filtering framework designed to optimize genome sequencing analysis, with particular emphasis on non-coding regions. It was applied retrospectively to 153 cases with previously negative exome and genome sequencing results: pre-annotation filtering (conservation metrics, allele frequency thresholds, splicing prediction scores) precedes phenotype-based prioritization and manual review. The framework identified pathogenic or likely pathogenic variants in 4.6% of cases, strong variants of uncertain significance in 1.3% and additional candidate variants in 5.2%. Several deep-intronic variants were confirmed by RNA studies and segregation analyses.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

Preprint on medRxiv, not peer reviewed. A 4.6% yield in already negative cases is worthwhile, and the framework rightly targets the non-coding regions that standard analyses overlook. However, the abstract compares SMART filtering neither with another tool nor with a simple reanalysis, and specifies neither the thresholds used nor code availability: the share of the gain attributable to the framework itself is not measured.

Analysis by Dr Thibaut Benquey