Cancer Genomics and Diagnostics

This study uses discrete diffusion models to generate tumor phylogenies, finding that mid-scale models outperform deeper ones and that diverse training improves cross-regime generalization.

TumorLens detects genetic and epigenetic alterations in cancer using long reads, revealing key somatic events and methylation patterns across tumor samples.

This study identifies elevated YBX1 mRNA expression as a marker of genomically unstable and clinically aggressive cancer across multiple tumor types.

The paper analyzes subclonal gene dosage effects in cancer using DNTR-seq, finding additive and compensatory transcriptional impacts and identifying transient clonality in ovarian and soft tissue sarcomas.

The paper introduces a graph-based framework to construct mutation-centric networks using long-range genomic interactions, demonstrating its ability to cluster cancer subtypes and prioritize non-coding driver mutations.

The paper identifies stop-loss mutations in cancer genomes, showing enrichment in cancer-related genes and potential immunological consequences.

This study shows that elevated cfDNA and fragment shortening in cancer are due to saturated hepatic clearance, not tumor burden alone.

ROCIT classifies tumor or non-tumor origin of reads from bulk long-read sequencing with high accuracy, improving somatic variant calling in tumors.

MuAt2, a Transformer model, classifies tumour types and subtypes from somatic variants, improving accuracy and calibration across cancer types and cohorts.

This paper introduces SPICE, a framework to infer copy-number events and model selection in cancer genomes, revealing 460 loci under selection and expanding the known cancer-related loci.

Arborist improves tumor phylogeny reconstruction by integrating bulk DNA-seq and low-pass scDNA-seq data, outperforming existing methods on simulated and biological datasets.

LRSomatic is a pipeline for somatic variant calling in long-read sequencing data, achieving state-of-the-art performance on cancer samples and resolving complex variants.

The paper compares germline and somatic variants in 40 tumor suppressor genes, finding distinct mutational patterns and locations shaped by different selection pressures.

This paper introduces omlta, an algorithm for aligning clonal trees to compare tumor progression, and applies it to cancer datasets.

POTTR identifies recurrent mutation trajectories in cancer and development using posets, resolving uncertainty in phylogenies and discovering significant new patterns.

The paper shows that cancer cells generate non-random, high-frequency mutations during drug adaptation, enriched near transcription factor binding sites.

The paper trains machine learning models to predict cancer driver mutations from tumor microenvironment composition across multiple cancers and validates them externally.

This study identifies cooperative and suppressive interclonal interactions during early Ras-driven tumourigenesis using a Drosophila model, revealing key modifiers like SWI/SNF and cell competition regulators.

The paper identifies novel markers for CTC subtypes and introduces CTCeek, a tool to annotate bona fide CTCs from scRNA-seq data.

APEX infers cancer gene expression from plasma cfChIP-seq data, outperforming existing methods and enabling subtype classification and biomarker identification.

OncoBERT models somatic mutations in cancer to identify patient-specific subtypes and improve prediction of treatment response by integrating mutational context with clinical biomarkers.

This study develops an ex vivo drug sensitivity testing platform for rare solid tumors and finds that high in vitro sensitivity correlates with clinical benefit and progression-free survival.

The paper analyzes two-hit inactivation frequencies in tumor suppressor genes across 9,000 tumors, identifying factors like mutation impact, zygosity, and chromosomal context that influence these frequencies.

The paper develops a framework to predict mutation effects on transcription factor binding using mutational signatures in breast cancer genomes, revealing subtype-specific rewiring of gene regulatory networks.

This paper develops DeepOmicsFFPE-PLUS, a transformer-based tool to accurately detect somatic variants from FFPE-derived WES and WGS data, outperforming existing methods and restoring biologically relevant mutational signatures.

ReDeNovo infers both known and novel mutational signatures from cancer genomes, outperforming existing methods and identifying new signatures in real data.

The paper identifies mutational signatures in cancer genomes that alter protein sequence motifs in signaling networks, linking mutation etiology to functional consequences.

This study shows that single-spun plasma is sufficient for accurate cfDNA analysis, challenging the need for double centrifugation in liquid biopsy workflows.

The paper introduces RePrint, a method to infer shared DNA repair mechanisms between mutational signatures, outperforming existing methods and enabling guilt-by-association predictions.

This paper identifies molecular features of monoclonal antibodies that enhance ctDNA detection by increasing its concentration in blood.