Single-cell RNA-seq and machine learning distinguish gain- and loss-of-function CASR variants

Classifying variants in genes where gain-of-function (GOF) and loss-of-function (LOF) mutations cause clinically distinct disorders poses a particular challenge for standard variant interpretation frameworks. A preprint posted to bioRxiv on 29 September 2026 describes a new approach applied to CASR, which encodes the calcium-sensing receptor: GOF variants in CASR cause autosomal dominant hypocalcaemia, whilst LOF variants cause familial hypocalciuric hypercalcaemia and neonatal severe hyperparathyroidism.

The authors engineered CASR-depleted HEK293 cells to express individual exogenous variants, then used single-cell RNA sequencing to capture transcriptional profiles for each variant. A supervised machine-learning classifier trained on these profiles was able to separate GOF from LOF variants with high accuracy, and also identified variants of uncertain significance (VUS) as more likely to be GOF or LOF based on their transcriptional signatures.

The study represents an advance for multiplexed assays of variant effect (MAVEs) — high-throughput experimental approaches that can characterise many variants simultaneously. The authors note that existing MAVE approaches often struggle with multi-mechanism genes, and argue that cell-type-resolved transcriptional readouts may offer a more granular signal than functional assays that collapse to a single output.

This is a preprint and has not yet undergone peer review. The findings should be interpreted as preliminary. The work was posted to bioRxiv (doi: 10.64898/2026.09.26.754690).