Computational predictions are not clinical evidence, researchers warn in CYLD variant dispute

A newly published comment in Genomics & Informatics argues that an earlier study of variants in the human CYLD gene overstepped what computational evidence can actually show, presenting predicted effects as established disease associations.

The comment, authored by Dimitri Tchernitchko and published as a Matters Arising item on 8 October 2026, responds to Roy et al., whose original article appeared on 31 May 2024. That study reported an extensive computational analysis of nonsynonymous single-nucleotide variants in CYLD and characterized several of them as potentially disease- or cancer-associated.

According to the comment, such analyses can be valuable for generating hypotheses, but computational predictions should be clearly distinguished from clinical or functional evidence. The central concern is that readers may interpret in silico predictions as if they carried the same weight as laboratory or clinical findings.

The key factual point raised is that among the 18 “high-risk nsSNPs” prioritized by the original authors, only one variant, p.Glu747Gly (E747G), has established clinical evidence supporting pathogenicity. The remaining variants, the comment states, lack sufficient evidence for pathogenic classification and should be considered variants of uncertain significance pending further evidence.

The authors highlight three major concerns with the original analysis. The first is the lack of independence among some of the prediction tools used for variant prioritization.

The third concern is the presentation of variants with no established clinical or experimental evidence as disease-associated. Under the ACMG/AMP framework for variant classification, the comment notes, computational evidence can provide supporting evidence but cannot substitute for clinical, population, segregation, or functional evidence.

In light of these concerns, the comment suggests that the terminology and conclusions of the original article be revised to distinguish computationally predicted effects from demonstrated disease associations. The recommendation is framed as a matter of how findings are described rather than a rejection of computational methods themselves.

Subject of Research: Over-interpretation of in silico predictions as clinical evidence in CYLD variant classification

Article Title: The risk of over-interpreting in silico predictions as clinical evidence: a comment on Roy et al. (2024)

Article References: Tchernitchko, D. (2026). The risk of over-interpreting in silico predictions as clinical evidence: a comment on Roy et al. (2024). Genomics & Informatics, 24(1), Article 18. https://doi.org/10.1186/s44342-026-00082-7

Image Credits: AI Generated

DOI: 10.1186/s44342-026-00082-7

Keywords: CYLD, in silico prediction, variant classification, ACMG/AMP, variants of uncertain significance, genomics, Matters Arising, pathogenicity, nsSNPs, Genomics & Informatics, risk, over-interpreting