Cancer drug sensitivity prediction from routine histology images

npj Precision Oncology, Journal Year: 2024, Volume and Issue: 8(1)

Published: Jan. 6, 2024

Abstract Drug sensitivity prediction models can aid in personalising cancer therapy, biomarker discovery, and drug design. Such require survival data from randomised controlled trials which be time consuming expensive. In this proof-of-concept study, we demonstrate for the first that deep learning link histological patterns whole slide images (WSIs) of Haematoxylin & Eosin (H&E) stained breast sections with sensitivities inferred cell lines. We employ patient-wise imputed gene expression-based mapping effects on lines to train a model predicts patients’ multiple drugs WSIs. show it is possible use routine WSIs predict profile patient number approved experimental drugs. also proposed approach identify cellular associated profiles patients.

Language: Английский

Assessing Genotype-Phenotype Correlations with Deep Learning in Colorectal Cancer: A Multi-Centric Study

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 8, 2025

Abstract Background Deep Learning (DL) has emerged as a powerful tool to predict genetic biomarkers directly from digitized Hematoxylin and Eosin (H&E) slides in colorectal cancer (CRC). However, few studies have systematically investigated the predictability of beyond routinely available alterations such microsatellite instability (MSI), BRAF KRAS mutations. Methods Our primary dataset comprised H&E CRC tumors across five cohorts totaling 1,376 patients who underwent comprehensive panel sequencing, with an additional 536 two public datasets for validation. We developed DL model using single transformer multiple slides. The model’s performance was compared against conventional single-target models, potential confounders were analyzed. Findings multi-target able numerous pathology slides, matching partly exceeding transformers. Area Under Receiver Operating Characteristic curve (AUROC, mean ± std) on external validation was: (0·78 0·01), hypermutation (0·88 MSI (0·93 RNF43 (0·86 0·01); this biomarker mirrored metrics co-occurrence analyses. high AUROCs largely correlated MSI, predictions depending considerably MSI-associated morphology upon pathological examination. Interpretation study demonstrates that transformers can status their pre-dictability is mainly associated phenotype, despite indications slight biomarker-inherent contributions phenotype. findings underscore need analyze AI-based oncology biomarkers. To enable this, we validated applicable other cancers larger, diverse datasets. Funding German Federal Ministry Health, Max-Eder-Programme Cancer Aid, Education Research, Academic Exchange Service, EU.

Language: Английский

Deep learning-based multimodal spatial transcriptomics analysis for cancer

Advances in cancer research, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 38

Published: Jan. 1, 2024

Language: Английский