Abstract B046: Filtering artifactual signal for DNA-methylation arrays in pediatric tumors: A benchmarking of preprocessing algorithms

Abstract Introduction: Methylome analysis holds promise in cancer research, enabling a better understanding of tumor biology and epigenetic regulation. Problem: Array-based methylation profiling can be compromised by recurrent somatic genetic alterations in cancer, which can cause nonspecific (off-target) hybridization of methylation probes, resulting in inaccurate measurements. The optimal preprocessing method to mitigate such artifacts in pediatric cancers remains unclear. We benchmarked two widely utilized pipelines to identify the most effective filtering approach in the context of pediatric tumors. Method ology: We analyzed 260 pediatric tumors profiled by MethylationEPIC arrays and matched whole-exome sequencing (WES). Raw methylation signals were preprocessed and transformed to beta-values using two pipelines: Minfi-ENmix and SeSAMe. SeSAMe suppresses signal noise at the individual probe level, whereas ENmix evaluates probe distributions across all samples to filter out probes with abnormal distribution. We listed from WES dataset the somatic alterations classified by copy number alterations (CNAs), single-nucleotide variants (SNVs), and insertions/deletions (Indels). We compared each pipeline’s sensitivity by quantifying the proportion of probes located within 10 base pairs of the alteration sites that were filtered. To assess the specificity, we used onco-heatmaps to visualize probes filtered or retained in six highly mutated regions. Results: The two pipelines showed over 85% concordance (difference in beta value < 0.1). Across all samples, 35,154,020 alterations were detected (CNAs were the most frequent alterations). On these sites, ENmix flagged 8.2% of probes as non-assessable (NAs), compared to 1.7% by SeSAMe. Of the probes filtered by SeSAMe, 80% overlapped with ENmix, while only 16% of ENmix-filtered probes were consistent with SeSAMe. In regions with two copies loss (CNA-2), ENmix filtered 7.3% probes versus 3.5% by SeSAMe, for one-copy loss (CNA-1), 6.9% vs. 1.1%; and for copy gain, 7.1% vs. 1.1%, respectively. ENmix also filtered more probes near SNVs and Indels (8.2–9.6%) than SeSAMe (2.0–7.0%). In highly mutated genes, ENmix tended to remove probes across all samples. In contrast, SeSAMe was more accurate in filtering probes from samples with corresponding genomic alterations, resulting in more precise filtering and higher specificity. Conclusion: ENmix demonstrated a higher sensitivity by broadly filtering non-specific signals on sequences with genomic alterations. SeSAMe, by contrast, offered greater specificity through conservative sample-specific filtering. Both pipelines provide valuable preprocessing strategies for methylome analysis in pediatric cancers, with the choice depending on the specific research objective, whether prioritizing sensitivity or specificity. Citation Format: Mona Patoughi, Stéphanie Bianco, Anas Belaktib, Charles Joly-Beauparlant, Thibault Mallevaey, Sylvie Langlois, Alex Richard-St-Hilaire, Thomas Sontag, Noel Raynal, Thai Hoa Tran, Daniel Sinnett, Arnaud Droit, Raoul Santiago. Filtering artifactual signal for DNA-methylation arrays in pediatric tumors: A benchmarking of preprocessing algorithms [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Discovery and Innovation in Pediatric Cancer— From Biology to Breakthrough Therapies; 2025 Sep 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_2):Abstract nr B046.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.