Computational Biology for Ecosystem and Human Health
Computational biology research in the Song group is directed toward developing mathematical models and multi-omics integration tools for predicting 1) metabolic behaviors of microbial and human cells and 2) interspecies interactions and microbial community dynamics in environmental and human microbiomes. For enhanced prediction, the group synergistically integrates process-based and data-driven modeling with ecological theory. Current projects include:
- Soil microbiome modeling
- Biogeochemical modeling
- Human microbiome modeling
- Human cell modeling
- Computational drug discovery
Soil Microbiome Modeling
- A grand challenge in soil microbiome research is to establish a molecular understanding of how the shifts in environmental conditions such as moisture contents and nutrient inputs reshape microbial interactions and community functions, and ultimately the collective metaphenome of the soil microbiome. Achieving this goal requires the development of predictive mathematical models that enable linking environmental conditions, microbial interactions, and microbiome functions. To this end, we combine genome-scale metabolic network modeling, agent-based modeling, the cybernetic approach, and machine learning to understand/predict condition-specific biochemical signatures in soil microbiomes, microbial interactions in chitin-degrading communities, and spatial assembly facilitating microbial interactions.
- Key collaborators: Soil Microbiome Scientific Focus Area (SFA) team at Pacific Northwest National Laboratory (PNNL) (Leads: Dr. Janet Jansson and Dr. Kirsten Hopfmockel), and KBase team at Argonne National Laboratory (ANL) (Lead: Dr. Christopher Henry)
Biogeochemical Reaction Modeling
- Our goal is to develop and demonstrate a new concept of biogeochemical reaction network modeling that accounts for fundamental microbial processes and associated metabolic principles. Major topics include: 1) modeling of historical contingency in biogeochemistry, 2) prediction of priming effect caused by groundwater and surface water mixing, 3) incorporation of high-resolution mass spectrometry data into biogeochemical modeling, and 4) development of a modeling pipeline from omics-informed biogeochemical and reactive transport modeling.
- Key collaborators: Subsurface Biogeochemical Research(SBR) team at PNNL (Leads: Dr. Timothy Scheibe, Dr. James Stegen, Dr. Xingyuan Chen, and Dr. Emily Graham), IDEAS-Watersheds project (Lead: Dr. David Moulton; PNNL-side lead: Dr. Xinyuan Chen) and KBase team at Argonne National Laboratory (ANL) (Lead: Dr. Christopher Henry)
Human Microbiome Modeling
- The overarching goal in our human microbiome modeling is to reveal the mechanistic relationships between the structure and function of the gut microbiome and the host genome and stress conditions. Major focuses in the current projects include metagenome-based disease prediction using machine learning techniques and modeling the dynamic shifts in microbial interactions in the gut microbiome in response to nutrient fluctuations.
- Key collaborators: Dr. Jennifer Auchtung at Nebraska Food for Health Center and Dr. Stephen Lindemann at Purdue University
Human Cell Modeling
- Our general interest is to develop computational tools for integrating multi-omics data into human metabolic network models to predict the shifts in metabolic pathways under diet, disease, and stress conditions, and ultimately to reveal a mechanistic linkage with the structural and functional changes in the gut microbiome. The current project focuses on identifying the beneficial and detrimental roles and the action of mechanisms of copper (Cu) in adipocyte functions, lipid metabolism, and the onset and progression of metabolic disorders.
- Key collaborators: Dr. Jaekwon Lee and Dr. Seung-Hyun Ro in the Department of Biochemistry at UNL
Computational Drug Discovery
- Drug development is a lengthy, complex, and costly process that could greatly be facilitated by computers. We use a suite of advanced data-driven modeling approaches including deep learning to (1) identify chemical motifs that control drug activity and DMPK (drug metabolism and pharmacokinetic)/ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties, and (2) suggest a narrow set of new drug candidates.
- Key collaborators: Dr. Hasoo Seong at Korea Research Institute of Chemical Technology (KRICT) in Korea