I. Integrative analysis of ’omics data to obtain insights into colon cancer-associated compositional shifts

The association between colon cancer and gut microbes is an intriguingly complex relationship, with no single microbe or pathogen, or microbial function, appearing to be causal.  Instead, colon cancer has been repeatedly linked to the overall gut ecology.  This suggests that this disease may be best studied by taking into account many different types of measurements as the gut microbial ecology changes during the onset and progression of colon cancer.  We are currently using an extensive, multi-omic profiling approach (e.g. meta-genomic, meta-transcriptomic, stool metabolome) to uncover the biomolecular and taxonomic features of colon cancer microbiome, with a goal to identify robust, universal signatures.  These signatures will span various biomolecular types and microbial taxa and their functions, as it is highly unlikely that only a small number of covariates portray the complex picture of colon cancer.

Our analysis entails the integration of all promising, but seemingly disparate candidate markers (ranging from particular taxa, functional groups, and biomolecules to microbe-microbe metabolic relationships) into a global, mechanistic framework.  Successful demonstration of this multi-omics approach will shift microbiome studies away from investigating only one data type (which is commonly seen in the literature), and towards a more systems-level approach contingent upon a thorough understanding of the entire gut ecosystem.

II. Developing statistical inference and machine-learning tools to non-invasively monitor health and disease

Large-scale, data-driven approaches to healthcare have the potential to address some of the most critical biomedical concerns, such as early disease diagnostics, assessment of therapeutic regimen efficacy, and real-time monitoring of health and disease.  Emerging technologies for personal data metrics and analytics, paired with non-invasive diagnostic methods, can be used to identify actionable biomarkers to guide clinical decision-making.

As a step towards the discovery of early indicators for disease disposition using gut microbiome, we aim to develop the computational approaches necessary to identify combinations of microbial taxa and/or biomolecules from patient stool samples that show strong association with a particular disease.  If there are indeed particular features in the gut, at concentrations that can be accurately and consistently detected in stool samples, then there is the very exciting possibility that gut pathologies, such as colon cancer, can be non-invasively detected at its early stages.  Moreover, stool samples are easily accessible (in contrast to tissue biopsies) for obtaining gene, transcript, protein, or metabolomic profiles.  As mentioned above, it is critical that many data types are monitored; this will mitigate confusion from false positives/negatives that occur by looking at only one data type.  And certainly, abnormalities in more data types are more credible than those in only one (hence, a good way to control for specificity).  We are working towards the realization of this potential to make patient stool a powerful window into gut wellness.

III. Network modeling of microbial community-scale metabolism to investigate global functional properties

Ultimately, gut microbiome analyses will evolve beyond descriptive, profiling investigations towards more hypothesis-driven, mechanism-focused studies.  However, progress in this direction will be contingent upon the maturation of our general understanding of the global inner workings of a microbial community.  Especially, we will need a deeper fundamental understanding of how biological communities in our gut organize, integrate, and carry out complex biochemical processes, and how these processes determine interaction relationships between microbial species.  Moreover, to realize the full potential of microbiome-based therapy via either targeted intervention or whole microbiome transplantation, a comprehensive computational framework is required for directing such manipulations and offering promising intervention routes.  Thus, there is an imperative need to develop, and continuously refine, community-level interaction models of gut microbial species, as well as to incorporate into these models information on biochemical mechanisms.

Our emphasis on community-scale modeling enables us to globally characterize and make predictions regarding the behavior of an entire microbiome community, and possibly offer details into the functional capabilities of the community.  Indeed, by incorporating individualized multi-omic measurements (mentioned above), this direction can lead to intriguing possibilities of dynamically monitoring one’s gut microbiome at the level of an entire ecosystem.  Eventually, we will be able to establish links between the patient, disease class, molecular signatures, and functional modules within a patient’s own microbial community.