Disease Prediction by Detecting and Integrating Connectomic Networks and Marginally Weak Signals

Many contemporary studies use individual genomic or imaging profiles for early prediction of cancer or neuropsychological outcomes, such as cancer subtypes and Alzheimer's disease stages. Current approaches ignore the connection structures of the genome and the brain (e.g. gene pathways or brain networks). Many genetic and imaging markers, despite having marginally weak effects, may exude strong predictive effects once considered together with their connected biomarkers. To find such weak signals, the inter-feature connectomic structure of the genome or brain must be explored first. However, given the ultrahigh-dimensional characteristic of genomic/neuroimaging profiles, identifying the whole genome/brain connectomic features is computationally prohibitive. This is also an impediment to detecting weak signals. In this work, we hypothesize that a large portion of the predictiveness of disease outcomes attributes to inter-marker connections as well as marginally weak signals. By detecting and integrating them, prediction accuracy can be significantly improved. We develop novel statistical/machine-learning algorithms for detecting network-based biomarkers for cancer or AD-related outcome prediction. The identified network signatures and weak signals will also enhance our understanding of the underlying mechanisms of disease development and progression.