Dissecting Response to Cancer Immunotherapy by Applying Bayesian Network Analysis to Flow Cytometry Data

Abstract

Cancer immunotherapy, specifically immune checkpoint blockade, has been found to be effective in the treatment of metastatic cancers. However, only a subset of patients achieve clinical responses. Elucidating pretreatment biomarkers predictive of sustained clinical response is a major research priority. Another research priority is evaluating changes in the immune system before and after treatment in responders vs. nonresponders. Our group has been studying immune networks as an accurate reflection of the global immune state. Flow cytometry (FACS, fluorescence-activated cell sorting) data characterizing immune cell panels in peripheral blood mononuclear cells (PBMC) from gastroesophageal adenocarcinoma (GEA) patients were used to analyze changes in immune networks in this setting. Here, we describe a novel computational pipeline to perform secondary analyses of FACS data using systems biology/machine learning techniques and concepts. The pipeline is centered around comparative Bayesian network analyses of immune networks and is capable of detecting strong signals that conventional methods (such as FlowJo manual gating) might miss. Future studies are planned to validate and follow up the immune biomarkers (and combinations/interactions thereof) associated with clinical responses identified with this computational pipeline.

Synthetic data generation with probabilistic Bayesian Networks

“…Much of the work detailed in this communication was spurred by our ongoing collaborative immuno-oncology study that involves comparative BN analyses of multi-dimensional FACS (fluorescence-activated cell sorting) and other immuno-oncology datasets obtained from patients with gastrointestinal and breast cancers undergoing checkpoint blockade immunotherapy treatments [ 3 ]. In that study, we aim to construct and compare BNs representing immune network states in sickness and health, before and after therapy, in responders and non-responders.…”

“…This activity is also known as causal discovery (inference) in graphical models. Bayesian Networks (BNs)-based dependency modeling is an established computational biology tool that is rapidly gaining acceptance in the diverse areas of biomedical data analysis, ranging from molecular evolution [ 1 ] to chromatin interactions [ 2 ] to flow cytometry [ 3 ] to diagnostic and prognostic clinical applications [ 4 , 5 ] to genomics [ 6 – 9 ] to transcriptomics [ 10 – 12 ] to metabolomics [ 13 ] to signaling [ 14 ] to microarray data analysis [ 15 ] to general multi-omics [ 16 – 19 ]. Comprehensive treatments of BN methodology can be found in textbooks [ 20 – 23 ] and review papers [ 24 – 27 ].…”

Editorial: Systems Biology Methods in Computational Immuno-Oncology

“…For example, there is only a limited concordance between the different network modeling methods in the scRNA-seq gene regulatory network space, and many of the network modeling methods are unsuccessful in recovering the whole "ground truth" (Chen and Mar 2018;Pratapa et al, 2020). Likewise, "conventional" immuno-oncology data analysis methods might miss strong signals due to the violation of linearity, normality, and other assumptions (Rodin et al, 2021). We believe that the solution lies in 1) development and application of the variety of methods, with different explicit and implicit assumptions, and 2) concentrating on, and highlighting, smaller sub-networks of interest-such networks can be more easily validated and refined in silico, thus partially sidestepping the "dimensionality curse".…”

Bayesian networks elucidate complex genomic landscapes in cancer

“…Bayesian networks have also been used in a number of varied cancer and genomic settings. For example Wang et al reconstruct regulatory networks using BNs 19 , Rodin et al build networks from flow cytometry data in the context of cancer immunotherapy 20 and Howey et al use BN in Mendelian randomization in the context of genetic epidemiology 21 . The related literature includes conjunctive BNs 22 , 23 which were originally introduced in the context of HIV but also proposed for cancer, direct learning of logic formulae 24 and BNs for fitness landscapes 25 .…”