Bioinformatics, Machine Learning, Systems Biology of Cancers

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

The goals of this project are: 1) Integration of mathematical modeling and bioinformatics to decipher the complex biology of pediatric cancers; 2) Model how DNA copy number alters expression of RNA; 3) Model how global methylation profiles alter expression of RNA; 4) Model how DNA copy number alters expression of protein; 5) Model how global methylation profiles alter expression of protein; 6) Model interactions of DNA copy number, messenger RNA (mRNA), and protein with the cancer phenotype; 7) Investigate pathways activated or suppressed in pediatric cancers; 8) Model and predict how can mutations lead to alteration of function and pathways; 9) Epigenetic profiling of rhabdomyosarcoma; 10) single-cell RNAseq and spatial proteomics and transcriptomics; 11) Perform machine learning and convolutional neural networks of H&E images to predict mutations and biological behavior of cancer. It has become increasingly clear that individual genes do not act in isolation within a cell, but acts in concert with other genes within pathways. In addition, many genes within a pathway are redundant such that many genes can perform a particular function, and conversely a single gene can have many functions depending on its cellular context. High throughput genome wide approaches allow the investigation of the complexity of cell as a whole. The integration of genomics and proteomics with functional analysis will be referred to as systems biology. My lab has utilized these systematic approaches to understand more clearly the biology of high-risk pediatric cancers. Systems biology seeks to integrate high-throughput biological studies to understand how the whole biological systems function. By studying the relationships and interactions between various parts of the biological system (NB in our case), including DNA copy number, methylation patterns, mRNA, and protein levels, cell growth, clinical parameters (age, stage and outcome (survival)), it is hoped that eventually this will enable a more complete understanding of pediatric cancers which will lead to improved survival of patients with minimal long term morbidity. The genome and proteome of a cell is a complex interrelated dynamic system. DNA copy number can impact the mRNA, mRNA is transcribed into proteins, and proteins control transcription by its action on DNA, and mRNA and proteins. We are using deep learning with convolutional neural networks to identify tumor features in H&E images that are associated with multi-omic data, including DNA, mRNA, and protein alterations, to increase the understanding of the behavior of cancers, predict diagnosis and outcome and facilitate precision medicine.