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

• Activity Number: 438 - Modern Statistical Learning Methods for High-Dimensional Biomedical Data: Treatment Heterogeneity and Data Integration
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Yanming Li* and Fengwei Yang and Xueping Zhou and Devin C Koestler and Wei Chen
• Companies: University of Kansas Medical Center and University of Kansas Medical Center and University of Pittsburgh and University of Kansas Medical Center and University of Pittsburgh
• Keywords: genome connectomic structure; Lung cancer prediction; Local gene network; Ultrahigh-dimensional classification; Survival

Abstract

Many contemporary studies use individual genomic profiles for early prediction of cancer outcomes, such as cancer subtypes and survivals. Current approaches base prediction using only biomarkers that are strongly correlated with the outcome. The connection structures of the genome are ignored in such marginal approaches. Many genetic biomarkers, despite having marginally weak effects, may exude strong predictive effects once considered together with their connected markers. Such signals are not detectable by themselves. In order to find them, the inter-feature connectomic structure has to be explored first. However, identifying the untralhigh-dimensional connectomic structure is computational prohibitive. This is also an impediment for detecting weak signals.

In this work, we develop a novel statistical/machine-learning algorithm for detecting connectomic genomic networks for lung cancer prediction. By detecting and integrating the connectomic structures of the genome and weak signals, accuracy of prediction can be significantly improved. The identified network or pathway signatures will also enhance our understanding on the mechanisms of cancer development and progression.