BioTwin
TwinMe
Sign in Store

BioTwin publishes medical research, clinical development, and wellness education. Clinical claims apply only where authorized. TwinMe wellness content is not medical advice, diagnosis, screening, treatment, or disease monitoring.

Publications Preprint 4 min

Metabolomic Profiling of Dried Blood Spots for Breast Cancer Detection: A Multi-Classifier Validation Study in 2,734 Participants

Preprint by Anctil, Hauguel, Rhéaume, Grobmyer, and Noel showing that untargeted metabolomics from a single dried blood spot detects breast cancer across 2,734 participants, with performance that is robust across six classifier families.

Nicolas Anctil, Pierrick Hauguel, Caroline Rhéaume, Stephen Grobmyer, Louis-Philippe Noel ·
oncologybreast cancermetabolomicsdried blood spotspreprint

Why it matters

Demonstrates that BioTwin's same-lab dried blood spot LC-MS protocol, the one validated for individual identification, carries enough signal to detect breast cancer at scale, with batch-aware validation that avoids the optimistically biased estimates common in the field.

View publication → Download file →

Summary

A cohort of 2,734 participants (114 biopsy-confirmed breast cancer cases and 2,620 non-cancer controls) was profiled by untargeted LC-MS/MS from a single dried blood spot (DBS), using a Thermo Scientific Orbitrap IQ-X coupled to a Vanquish UHPLC system. A pre-specified panel of 39 metabolites meeting MSI Level 1 identification criteria, selected on an independent prior workflow and applied here unchanged, was evaluated with six standard supervised-learning architectures (LASSO, Elastic Net, Linear SVM, PLS-DA, OPLS-DA, XGBoost). Detection performance was robust across classifier families, reaching an AUC of 0.928 for LASSO and 0.949 for XGBoost.

The paper addresses two recurring bottlenecks in blood-based cancer metabolomics: the field’s near-exclusive reliance on serum or plasma, which require venipuncture and cold-chain logistics, and machine-learning protocols that are blind to analytical batch structure and therefore report optimistically biased performance. DBS sampling removes the logistical barrier, and a batch-aware validation design keeps the reported estimates honest.

Why it matters

This is the disease-classification counterpart to BioTwin’s individual-identification work: the same minimally invasive, self-collected DBS matrix and the same single-laboratory LC-MS protocol, now applied to early breast cancer detection. Robustness across six independent classifier families, rather than a single tuned model, is what makes the signal credible.

Authors

Nicolas Anctil (1), Pierrick Hauguel (1), Caroline Rhéaume (2), Stephen Grobmyer (3), Louis-Philippe Noel (1). Corresponding author: Pierrick Hauguel (phauguel@biotwin.ai).

  1. Research and Development Department, BioTwin Inc., Quebec City, QC, Canada
  2. Université Laval, Quebec City, QC, Canada
  3. Oncology Institute, Fatima bint Mubarak Center, Cleveland Clinic Abu Dhabi, Abu Dhabi, United Arab Emirates

Competing interests: Nicolas Anctil, Pierrick Hauguel and Louis-Philippe Noel are employees and shareholders of BioTwin Inc. Caroline Rhéaume and Stephen Grobmyer are external academic and clinical collaborators.

Important: This article may discuss BioTwin research, medical vision, regulated clinical pathways, or TwinMe wellness education. TwinMe wellness outputs are not medical or laboratory tests. BioTwin clinical outputs are available only where authorized and through licensed healthcare professionals.