Advancing the science behind genomic technology

We investigate the scientific questions that shape how genomic technologies are built, evaluated and used.

We move the field forward through reproducible research, open methods and computational systems designed for real genomic work.

The records below cover three different questions. Evidence admissibility examines who authorises the reference snapshot used by a classifier. The real-world performance framework separates analytical, classification and clinical performance and demonstrates a method on 97 scored cases. The endometrial carcinoma study tests developmental regulon associations, robustness and cross-cohort concordance.

Each detail page includes the abstract, authors, publication date, version, DOI and evidence boundary. Follow the primary record to inspect the published article or preprint and, where available, its computational repository and archived release. A methodological proposal, a demonstration and an independently confirmed result carry different evidential weight. The pages preserve those distinctions and identify work that has not been peer reviewed.

03 public recordsPublished articles and preprintsOpen reproducibility material

Research

Published / Real-world performance

A Methodological Framework for Real-World Performance Studies of Clinical Variant Classification Platforms at Early Organizational Stages

A route between informal internal testing and formal regulatory adjudication, demonstrated on 97 scored cases.

Early-stage variant classification platforms need a performance method that is stronger than an internal comparison but feasible before a multi-adjudicator regulatory study. A single peer laboratory is difficult to treat as ground truth when qualified laboratories can disagree. Formal adjudication, however, requires people and quality infrastructure that a small developer may not yet have.

This is a methodological proposal with a single-platform demonstration. It uses one peer comparator, develops the taxonomy on the same cohorts used to illustrate it and carries conflicts of interest that require independent replication.

14 September 2026

Preprint / Clinical variant classification

Evidence Admissibility Is the Load-Bearing Human Gate in Auditable Clinical Variant Classification

Human authorization at evidence admissibility, before a version-locked snapshot enters deterministic classification.

Clinical AI oversight is often concentrated at final output. In variant classification, an autonomous component can affect the decision earlier by changing the reference data and evidence snapshot consumed by the classifier.

The framework addresses auditability and governance. It does not establish classification correctness, clinical validation, comparative effectiveness or fitness for purpose. Complete separation also increases the workload of human literature curation.

14 July 2026

Preprint / Precision oncology

Robustness and cross-cohort concordance of developmental regulons in endometrial carcinoma

Association, target-set robustness and transportability tested across TCGA-UCEC and CPTAC-UCEC.

The analysis asks whether developmental regulon signals in endometrial carcinoma survive changes in target composition and retain their direction in an independent cohort. It keeps pointwise association, target-set robustness and cross-cohort transportability as separate claims.

The results support directional concordance for selected regulons. They do not establish causal transcription-factor activity, biomarker validity, therapeutic dependence, treatment response or clinical utility. The preprint has not been peer reviewed.

2 August 2026

Find us on