A study published in Nature Medicine demonstrates that deep learning can predict microsatellite instability (MSI) directly from histology in gastrointestinal (GI) cancer. This method uses hematoxylin-eosin (HE) histology slides, which are widely available, to identify MSI, a key factor in determining whether patients will respond to immunotherapy. The research team developed a deep residual learning model, ResNet18, which achieved high accuracy in classifying MSI and MSS (microsatellite stable) in GI cancer tissues. The model was tested on data from The Cancer Genome Atlas (TCGA) and validated on an external dataset, the DACHS study, showing robust performance across different cancer types and tissue preservation methods. The model's ability to predict MSI without the need for additional genetic or immunohistochemical tests could significantly improve the identification of patients who may benefit from immunotherapy. The study highlights the potential of deep learning in oncology for more efficient and accessible biomarker detection, enabling broader application of immunotherapy. The method is cost-effective and can be implemented in tertiary care centers, offering a promising approach for improving patient outcomes in GI cancer.A study published in Nature Medicine demonstrates that deep learning can predict microsatellite instability (MSI) directly from histology in gastrointestinal (GI) cancer. This method uses hematoxylin-eosin (HE) histology slides, which are widely available, to identify MSI, a key factor in determining whether patients will respond to immunotherapy. The research team developed a deep residual learning model, ResNet18, which achieved high accuracy in classifying MSI and MSS (microsatellite stable) in GI cancer tissues. The model was tested on data from The Cancer Genome Atlas (TCGA) and validated on an external dataset, the DACHS study, showing robust performance across different cancer types and tissue preservation methods. The model's ability to predict MSI without the need for additional genetic or immunohistochemical tests could significantly improve the identification of patients who may benefit from immunotherapy. The study highlights the potential of deep learning in oncology for more efficient and accessible biomarker detection, enabling broader application of immunotherapy. The method is cost-effective and can be implemented in tertiary care centers, offering a promising approach for improving patient outcomes in GI cancer.