Research

Summary of Evidence Paper

AI-Assisted Genomic Variant Interpretation and Reclassification: A Summary of Evidence on SerenityGx

November 18, 2026

Authors: Shruti Chopra, PhD | Vandana Yadav, MS | Stan Kachnowski, PhD, MPA

A genetic test report reflects the evidence available on the day it is issued. That evidence keeps changing, and classifications change with it. Laboratory studies reviewed in this paper report that 12.4% of 1,209 BRCA1/2 variants (Mighton et al., 2019) and 7.2% of 8,075 variants (Hahn et al., 2024) were reclassified at one Ontario laboratory, and that 4.6% of more than 2 million variants were reclassified at a large commercial laboratory (Kobayashi et al., 2024). Variants of uncertain significance (VUS) are the most common starting point, and they resolve slowly: 65.5% of VUS first seen in 2015 were still unresolved when the Kobayashi study ended in 2023. Few laboratories have scalable infrastructure to re-review earlier reports or to connect a change to the patients affected.

SerenityGx is an AI-assisted genomic variant monitoring and review platform
that runs alongside a laboratory’s existing interpretation software. It checks
ClinVar, gnomAD, PubMed, and web search for each watch listed variant,
extracts facts with page and section citations, proposes ACMG/AMP criteria,
and runs ten deterministic code checks before a reviewer sees the result. It
manages three processes: Evidence Review, Reclassification, and Patient Notification. A classification changes only after two explicit human actions; override is restricted to Lab Directors with documented justification, and each action is recorded with the reviewer’s identity and the time. SerenityGx is clinical decision support software. It does not sequence samples, issue autonomous determinations or send information to patients, and the company describes it as a hosted service that is not a regulated product.

HITLAB finds a clearly defined, human-controlled and auditable design
addressing a well-documented gap,supported by early, formative evidence…