BC-Predict: Mining of signal biomarkers and production of models for early-stage breast cancer subtyping and prognosis

Sangeetha Muthamilselvan¹Natarajan Vaithilingam²Ashok Palaniappan^1*

• ¹Systems Computational Biology Lab, School of Chemical and Biotechnology, SASTRA University, Thanjavur, India
• ²Lincoln City Hospital, United Lincolnshire Hospitals NHS Trust, Lincoln, United Kingdom

Disease heterogeneity is the hallmark of breast cancer, which remains the most common female malignancy. With a consistent increase in mortality and disease burden, there exists a need for effective early-stage theragnostic and prognostic biomarkers. In this work, we improved on BrcaDx (https://apalania.shinyapps.io/brcadx/) for cancer vs control screening and examined a cluster of adjoining learning problems in breast cancer heterogeneity: (i) identification of metastatic cancers; (ii) molecular subtyping (TNBC, HER2, or luminal); and (iii) histological subtyping (invasive ductal or invasive lobular). We analysed the transcriptomic profiles of breast cancer patients from public-domain databases such as the TCGA using stage-encoded problem-specific statistical models of gene expression and unveiled a handful of stage-salient and progression-significant genes. Using a consensus approach, we identified potential machine learning features, and considered six model classes for each learning problem, with hyperparameter optimization on a training dataset and evaluation on a holdout test dataset. A nested approach enabled us to identify the best model class for each learning problem. External validation of the best models yielded balanced accuracies of 97.42% for cancer vs normal; 88.22% for metastatic v/s non metastatic; 88.79% for ternary molecular subtyping; and ensemble accuracy of 94.23% for histological subtyping. The model for molecular subtyping was also validated on a 26-sample TNBC-only out-of-distribution cohort, yielding 25 correct predictions. In addition, we performed a late integration of multi-omics datasets by validating the feature space used in each problem with miRNA profiles, methylation profiles, and commercial breast cancer panels. Pending prospective studies, we have translated the models into BC-Predict that forks the best models developed for each problem and provides an integrated readout of input instances of expression data, together with uncertainty estimates. BC-Predict is freely available for non-commercial purposes at:https://apalania.shinyapps.io/BC-Predict.