Nicholas Lee-Ping Chia

Nicholas received his Ph.D. from The Ohio State University in 2006, and from 2006-2011 was a postdoctoral fellow at University of Illinois Urbana-Champaign. He then went on to a Senior Researcher at the Institute for Systems Biology before becoming a faculty member at Mayo Clinic. He has served as junior editor for several journals, including mSystems and Frontiers. Selected honours include the Humboldt Fellowship and Fredrick P. Li Impact Award (AACR).

Some of my interests include:

• Inverse Reinforement Learning for Cancer Evolution: We develop multiple generative modeling framework for treating cancer that encodes insights of tumor development and progression on the molecular level, e.g., DNA, RNA, proteins, that describes the behavior of cancer. This can be used to extrapolate the evolutionary trajectory of any tumor and thereby forecast a tumor’s adaptive response to environmental exposures or treatments.
• Pre-training Methods for Representation Learning in Multi-Omics Data Using Large Language Models:The high dimensionality and heterogeneity of the multi-omics data--DNA, RNA, proteomics, methylation, microbiome--data pose challenges for analysis. Pre-training methods are commonly used in large language models to transform sequences of words into a topological space where similar words and sentences are embedded near each other. The goal of this project is to explore the effectiveness of pre-training methods for representation learning in multi-omics data using large language models.
• Personalized Probiotic Design: A probiotic’s ability to establish itself and carry out a desired function depends on the species-species interactions between the probiotic strain and an individuals’s microbial community. However, each individual’s starting gut microbiome is different. An automated, personalized approach to probiotic design---one that focuses on identifying which probiotics can become established and carry out the desired function in a given patient has the potential to propel the field forward. Starting only with sequence data for a patient’s gut microbiome, we build a pipeline that systematically identifies the most promising probiotic candidates by subjecting them to in silico testing.

Given the number of technologies employed around the primary goal of my lab, these have resulted in a number of side-projects that have or are close to having real-world impact.

• Answers-in-Hours: Quite simply put, this project uses sequencing to identify pathogenic organisms before they cause a surgical site infection. A simple approach with so-far strong results in the setting of pancreatic surgery that comes from lining up a number of technical achievements in sample processing and bioinformatics. Currently an open-enrollment FDA-approved clinical trial.
• Role of the Microbiome in Early Colorectal Cancer (CRC): Our work includes data-based studies of adenomas, CRC, and the microbiome. In addition, important advances in integrated multi-omics analyses focused on potential links between microbial metabolism and microbial metabolites.
• Foundation Model for Pathology: Histopathology images play a pivotal role in diagnosing and treating cancer. These images pose unique challenges due to their giga-pixel resolution, complex morphological phenotypes influenced by granular features like cellular substructures, and considerable inter- and intra-observer variability. These hurdles often hamper the development of dependable AI models. Current visual pre-trained models, generally trained on coarse-grained, object-centric images like those in ImageNet, have demonstrated limited success when applied to the field of pathology. Recognizing these limitations, one of my current research interests lies in creating a domain-specific foundation model tailor-made for pathology. This approach addresses the intricacies of histopathology images, ensuring that the model is precisely aligned with the realities of the data and the specific needs of the field. I believe such dedicated models will catalyze significant advancements in cancer diagnosis and treatment, optimizing patient care through more accurate and nuanced AI-driven insights.

chia [at] anl [dot] gov