Calibration of in-frame indel variant effect predictors for clinical variant classification

Area of Science:

• Human genetics
• Genomic variation
• Clinical bioinformatics

Background:

• Insertions and deletions (indels) are significant genetic variations with functional impacts.
• Short in-frame indels are understudied and difficult to interpret clinically compared to single nucleotide variants.
• Existing computational tools for missense variants are well-validated, but those for indels lack clinical utility certainty.

Purpose of the Study:

• To calibrate computational prediction tools for the clinical classification of in-frame indels.
• To establish evidence thresholds for pathogenic and benign indel classification based on ACMG/AMP guidelines.
• To assess the clinical utility and performance of current in-frame indel prediction tools.

Main Methods:

• Construction of a high-confidence dataset of in-frame indel variants (≤ 50bp) from clinical and population databases.
• Estimation of prior probabilities for pathogenicity of rare in-frame indels in disease genes.
• Application of a statistical framework using local posterior probabilities to set score thresholds for eight computational tools.

Main Results:

• All evaluated in-frame indel predictors achieved multiple evidence levels for pathogenicity and/or benignity.
• The calibrated tools demonstrate measurable clinical value in variant classification.
• Performance of indel predictors was consistently lower than that of missense variant predictors.

Conclusions:

• Computational tools for in-frame indel classification have been successfully calibrated, showing clinical utility.
• Despite demonstrated value, current indel predictors exhibit limitations compared to missense predictors.
• There is a clear need for the development of enhanced computational approaches for indel variant interpretation.