Deep Learning model accurately classifies metastatic tumors from primary tumors based on mutational signatures

Abstract

Metastatic propagation is the leading cause of death for most cancers. Prediction and elucidation of metastatic process is crucial for the therapeutic treatment of cancers. Even though somatic mutations have been directly linked to tumorigenesis and metastasis, it is less explored whether the metastatic events can be identified through genomic mutation signatures, a concise representation of the mutational processes. Here, applying mutation signatures as input features calculated from Whole-Exome Sequencing (WES) data of TCGA and other metastatic cohorts, we developed MetaWise, a Deep Neural Network (DNN) model. This model accurately classified metastatic tumors from primary tumors. Signatures of non-coding mutations also have a major impact on the model performance. SHapley Additive exPlanations (SHAP) and Local Surrogate (LIME) analysis into the MetaWise model identified several mutational signatures directly correlated to metastatic spread in cancers, including APOBEC-mutagenesis, UV-induced signatures and DNA damage response deficiency signatures.

Sparse Modeling of Genomic Landscape Identifies Pathogenic Processes and Therapeutic Targets in Metastatic Breast Cancer

“…We evaluated the performance of the updated approach and compared with our previous model, MetaWise 17 by five-fold cross validation. We measured the accuracy, recall, precision, specificity, F1score, the matthews correlation coefficient (MCC) and the area under the receiver operating characteristic curve (AUROC) and the area under the precision recall curve (AUPR) on both internal and external test sets.…”

“…AI-based prediction models have been developed based on clinical data, such as medical images, gene expression profiles, etc [9][10][11][12][13][14] . At the same time, deep learning architectures are also being developed for the prediction of metastasis, including multilayer perceptron, convolutional neural networks, autoencoders, etc [15][16][17][18] . These models aim to solve a binary classification problem by classifying samples as either metastatic or non-metastatic.…”

“…These models aim to solve a binary classification problem by classifying samples as either metastatic or non-metastatic. Several studies have shown the importance of mutational signatures in identifying cancer-associated genes and demonstrated some promising results on how these models guide precision medicine approaches 17,19 . However, major challenges remain to be addressed, including limited data accessibility, incorporation of multi-omics data, especially with extensive phenotypic annotations, limited generalizability across diverse patient cohorts, and the poor interpretability of DL models.…”

“…This can help to reduce overfitting, improve model interpretability, and decrease computational requirements. In this study, we introduce an updated and extended version of MetaWise 17 . In comparison to our previous version, the new MetaWise 2.0 benefits from pruning technology, and offers model the generalization capability for data from different cohorts.…”

Unveiling Pathogenesis and Therapeutic Targets of Metastatic Breast Cancer Through Sparse Modelling of Genomic Landscape

“…AI-based prediction models have been developed based on clinical data, such as medical images, gene expression pro les, etc [9][10][11][12][13][14] . At the same time, deep learning architectures are also being developed for the prediction of metastasis, including multilayer perceptron, convolutional neural networks, autoencoders, etc [15][16][17][18] . These models aim to solve a binary classi cation problem by classifying samples as either metastatic or non-metastatic.…”