Inference-based machine learning models to classify breast cancer subtypes

• Ion channels are promising molecules for classification of breast cancer subtypes • Inclusion of transcriptomic and methylomic data pertaining to ion channels could result in robust breast cancer subtype classification model • Inference-based machine learning models indicate overlaps of ion channels in subtypes with features identified using SHapley Additive exPlanations • Most likely, here is the foremost integrated ion channels multiOmics data-based machine learning framework to classify subtypes of breast cancer Breast cancer, a multifaceted disease, is classified into various subtypes. These subtypes are currently identified through immunohistochemistry or gene expression profiling, that in turn assist in determining appropriate treatment strategies. With the present therapeutics, breast cancer patients often undergo simultaneous side-effects during and after intense therapy, necessitating new biomarker-based targets. Ion channels regulate crucial cellular functions and can influence chemo-resistance and tumor growth. As significant drug targets, they offer promising therapeutic opportunities. Most recently, our group identified subtype alterations in ion channels and their regulatory elements in breast cancer, highlighting their potentiality for targeted treatment. Here, models based on support vector classifiers were developed, incorporating transcriptomic, methylomic, and genomic data. Using results from our previous analysis as features, these integrated models classified samples into molecular subtypes. The transcriptomic model could classify samples with a sensitivity of 0.63, specificity of 0.92 and area under the curve (AUC) of 0.77. The integrated methylomic-transcriptomic model could classify the same samples with a sensitivity of 0.71, specificity of 0.93 and AUC of 0.82 whereas the integrated genomic-methylomic-transcriptomic model resulted in a sensitivity of 0.63, specificity of 0.92 and AUC of 0.77. Utilising SHapley Additive exPlanations score calculations for features, the study further reports the ion channels in an order of importance taken into consideration by each model to classify each of the subtypes. Most likely, the study outlines a strategy for classifying breast cancer subtypes, emphasizing the potential role of ion channels in this classification, based on their observed alterations.