Causal DNA Methylation Biomarkers Discovery Approach against Regression-based Latent Space of RNA to Protein Expression Translation Process

The central dogma of molecular biology flows from DNA to RNA to Proteins. This paper studies the causal estimation of a given confounder, the DNA methylation biomarker, against the relationship in-between of its impacted dual successive multi-outcome, which is the translation process from RNA to Proteins expression. The translation process is extracted from a precisely designed Autoencoder, then the latent space is reduced to represent dual multi-outcome, the highly causal impact biomarkers are then utilized for normal/ cancer classification. The proposed approach is applied on CPTAC-3 Head & Neck(H&N), and Kidney cancer in comparison to the commonly LIMMA approach. The proposed approach provided better classification accuracy, however with minimum possible features, the highest reported accuracy 99.39% with 9 features for H&N cancer, while the LIMMA is 97.67% with 92,815 features, and Kidney 100% with 71 features while LIMMA is 100% using 70,587 features.