bims-fragic 2025-11-16 papers

Issue of 2025–11–16
three papers selected by
Laura Mannarino, Humanitas Research

Brief Bioinform. 2025 Nov 01. pii: bbaf599. [Epub ahead of print]26(6):

Early cancer detection via multi-omics cfDNA fragmentation using early-late fusion neural network with sample-modality evaluation.

Libo Lu, Yunze Wang, Xionghui Zhou.

Cell-free DNA (cfDNA) fragmentation patterns reflect epigenetic modifications and are promising biomarkers for early cancer detection. While integrating diverse fragmentomic signals can improve accuracy, high modality dimensionality, and limited samples challenge effective multimodal fusion. We present Early-Late fusion with Sample-Modality evaluation (ELSM), a two-stage neural network integrating 13 fragmentomic feature spaces with sample-wise modality evaluation to capture complementary signals. Across five datasets of 1994 samples from 10 cancer types, ELSM outperforms unimodal and advanced multimodal models for cancer detection and tissue-of-origin prediction, achieving an AUC of 0.972 for pan-cancer diagnosis and 0.922 in an independent gastric cancer cohort, with a median tissue-of-origin accuracy of 0.683. Analysis of key genomic regions identified by ELSM reveals robust interpretability aligned with known oncogenic pathways. ELSM provides a powerful and interpretable framework for integrative multi-omics analysis with strong potential for clinical translation in early cancer detection.

Keywords: cancer analysis; cell-free DNA; cohort study; machine learning; multimodal data processing

DOI: https://doi.org/10.1093/bib/bbaf599

Bioinform Adv. 2025 ;5(1): vbaf236

FinaleToolkit: accelerating cell-free DNA fragmentation analysis with a high-speed computational toolkit.

James Wenhan Li, Ravi Bandaru, Kundan Baliga, Yaping Liu.

Motivation: Cell-free DNA (cfDNA) fragmentation pattern represents a promising non-invasive biomarker for disease diagnosis and prognosis. Numerous fragmentation features, such as end motif and window protection score (WPS), have been characterized in cfDNA genomic sequencing. However, the analytical tools developed in these studies are often not released to the liquid biopsy community or are inefficient for genome-wide analysis in large datasets.
Results: To address this gap, we have developed FinaleToolkit, a fast and memory-efficient Python package designed to generate comprehensive fragmentation features from large cfDNA genomic sequencing data. For instance, FinaleToolkit can generate genome-wide WPS features from a ∼100× cfDNA whole-genome sequencing (WGS) dataset with over 1 billion fragments in 0.7 h, offering up to a ∼50-fold increase in processing speed compared to original implementations in the same dataset. We have benchmarked FinaleToolkit against original approaches or implementations where possible, confirming its efficacy. Furthermore, FinaleToolkit enabled the genome-wide analysis of fragmentation patterns over arbitrary genomic intervals, significantly boosting the performance for cancer early detection.
Availability and implementation: FinaleToolkit is open source and thoroughly documented with both command line interface and Python application programming interface (API) to facilitate its widespread adoption and use within the research community: https://github.com/epifluidlab/FinaleToolkit.

DOI: https://doi.org/10.1093/bioadv/vbaf236

Eur J Cancer. 2025 Nov 05. pii: S0959-8049(25)00983-9. [Epub ahead of print]231 116097

Characterization of the cell-free DNA released by tumor organoids derived from colorectal cancer patients.

Alexia Mirandola, Jérôme Cartry, Sabrina Bedja, Ekaterina Pisareva, Corinne Prevostel, Céline Gongora, Benjamin Ginter, Antoine Italiano, Michel Ducreux, Alice Boilève, Jacques R R Mathieu, Fanny Jaulin, Alain R Thierry.

INTRODUCTION: Patient-derived tumor organoids (PDTOs) have histological, molecular and clinical (drug sensitivity) characteristics comparable to those of their originating tumors. However, little is known about their ability to replicate the release of tumor-derived DNA.
METHODS: Supernatants from 21 colorectal cancer PDTO cultures, established from 13 patients, were prospectively collected. The presence, structure, and mutational landscape of nuclear (cf-nDNA) and mitochondrial (cf-mtDNA) cell-free DNA as well as extracellular mitochondria (ex-Mito) were analyzed using qPCR, fragmentomics and shallow whole-genome sequencing. Mutation profiling was performed via IntPlex qPCR and whole-exome sequencing (WES).
RESULTS: Cf-nDNA was detected in 95 % of PDTO supernatants with concentrations ranging from 0.009 to 209 ng/mL. Cf-nDNA fragment size analysis revealed patterns consistent with circulating DNA, including mononucleosome-associated profile. Cf-mtDNA was present in all samples (0.27-89.2 pg/mL) and extracellular mitochondria was also detected (0.009-17.4 pg/mL). A strong concordance (>85 %) was observed between oncogenic mutations in cfDNA and the molecular alterations detected in PDTOs and patient tumors.
CONCLUSION: PDTOs release both nuclear and mitochondrial cfDNA into their culture medium displaying high similarity with patient-derived circulating DNA (cirDNA), including fragmentation patterns and oncogenic mutations. This study strengthens the relevance of the PDTOs as patient tumors models and highlights the potential of analyzing PDTO-derived cfDNA as a non-invasive approach to investigate tumor evolution and as a valuable tool to support functional precision oncology.

Keywords: Cell-free DNA (cfDNA); Circulating DNA; Fragmentomics; Patient-derived tumor organoids (PDTOs)

DOI: https://doi.org/10.1016/j.ejca.2025.116097