Daniel De Carvalho
responses to interview questions from Tal Behar, Precision World Medicine Conference
responses to interview questions from Tal Behar, Precision World Medicine Conference
Daniel D. De Carvalho, PhD
Professor, University of Toronto; ALLAN SLAIGHT Senior Scientist, Princess Margaret Cancer Centre/University Health Network; Co-Founder and CSO, Adela
Professor, University of Toronto; ALLAN SLAIGHT Senior Scientist, Princess Margaret Cancer Centre/University Health Network; Co-Founder and CSO, Adela
- Where does methylation add indispensable value in MRD vs mutation‑only approaches (e.g., low‑shedding tumors, tissue‑of‑origin context, earlier relapse signal), and what prospective study would convince frontline oncologists?
Methylation adds value where mutation-only MRD struggles because it provides an extremely dense, lineage-informative signal that remains detectable at very low tumor fractions and carries tissue-of-origin context. Because of the large number of differentially methylated regions in cancer cells, methylations allow high sensitivity ctDNA detection even without use of tissue-informed personalized panels. That matters for low-shedding tumors, for relapses driven by subclones not captured in baseline tissue, for when tissue is not available, and when a faster result is necessary. A number of clinical studies have shown clinical validation of cfDNA methylation for several of these applications. Recently, we published a clinical validation of Adela’s tissue-agnostic genome-wide methylome enrichment assay for MRD detection in patients with Head and Neck cancer. The MRD detection test showed high sensitivity for identifying recurrence at high specificity across different anatomical sites, HPV status, and treatment regimens, highlighting the broad applicability of methylation for MRD detection (Liu et al, Annals of Oncology 2025).
- Which near‑term clinical indication for cfDNA methylome profiling will cross the evidence bar first (high‑risk screening, surveillance after curative therapy, or something else), and what endpoint (earlier stage shift, reduction in unnecessary procedures, survival surrogates) will matter most?
The first indication to cross that bar is post-curative surveillance, where prevalence is higher, follow-up pathways are standardized, and earlier molecular detection can trigger concrete clinical interventions. In this setting, methylation’s sensitivity at low burden, the easier clinical implementation of a tissue-agnostic assay will make a difference. High-risk screening will follow, but surveillance is positioned to demonstrate hard outcomes fastest.
- Looking ahead, what AI or multi‑omic combos (methylation + fragmentomics + mutations/proteins) are most likely to reduce false positives without sacrificing early‑stage sensitivity?
I’m very excited for the potential of AI-based algorithms for cfDNA methylome data. Since methylation captures intrinsic tumour biology, tissue-of-origin information and transcriptional profiles, cfDNA methylomes are information dense. Add this to cost-efficient assays that are broadly and dynamically capturing the cfDNA methylome, you then have a great substrate for AI-based algorithms. At Adela, we are building both the cfDNA methylome databases and AI-based algorithms to push clinical performance further.
- Where does Adela see the greatest opportunities for its platform across the cancer care continuum?
Adela’s platform is best deployed across a few points on the cancer continuum. Near-term, post-curative MRD, surveillance, and response monitoring are the clearest opportunities: sensitive, tissue-naïve methylome readouts detect response/relapse earlier integrating cleanly into existing follow-up schedules. In addition to therapy monitoring, response prediction will also benefit from methylome signatures that reflect tumor state, not just burden, yielding fast and more informative readouts than imaging alone. In the medium term, risk-adapted screening becomes feasible by combining methylation with AI-based algorithms to achieve the positive predictive value required for responsible population use.
