Our genome is the inherited record of what makes each of us unique. Personal genomes are large and complex, but they are readily measured and permanently associate with the individual. By comparison, human cells, tissues, and commensal organisms are more difficult to define. They show context-dependent, time- varying, and collective behaviors that complicate analyses at a population scale. Human disease states are even more challenging, in that it is often difficult or impossible to go back and examine the cell, tissue, or microbial physiology of the individual before the onset of disease. Connecting human variation to multi-cell function is fundamentally important to society in multiple ways. An explanatory and predictive understanding will empower individuals to appreciate how their genotype relates to potential phenotypes. It also will enable drug companies to identify precisely the patients most likely to respond to their therapies. Individualized “precision models” may one day expand the definition of wellness to include omics-scale inferences of biomolecular and cellular function.
Precision modeling has become feasible to consider as the most-relevant tools, resources, and innovations begin to converge. There are now many publicly available omics datasets on diverse human populations. Patient-derived cells are widely banked as human organoids or induced pluripotent stem cells (iPSCs). A path toward generalized predictions has been paved by systems-modeling approaches in bioengineering and machine-learning techniques in computer science. What is lacking is a convergent, trandisciplinary effort by a team of biologists, clinicians, omicists, informaticians, and engineers focused on a singular goal: precision multi-cell modeling.
Viewed as an engineered system, precision multi-cell models appear intractably complex, seeking to predict complex outputs from complex inputs with a difficult transfer function in between. The odds of identifying a single biomarker that directly maps to output (e.g., cholesterol → cardiovascular risk) are vanishingly small. Additionally, human-derived material will always be an imperfect avatar of the individual, regardless of that material’s technologic sophistication. Instead of aspiring to emulate the complexity of the transfer function, we seek to change the fundamental problem statement into a series of steps that are individually more manageable. In signal processing, this goal is achieved through reversible transforms to a domain in which the transfer function is easier to compute. Analogously, mechanistic systems models provide explanatory and predictive power to move back and forth between molecular-level variations and specific multi- cell phenotypes. Gaps in mechanistic detail can be filled by statistical and machine-learning approaches to achieve input-output predictions based on quantitative data. By leveraging computation in all its forms, standardized organoid or iPSC platforms become the transfer function for Precision In Silico-Cellular Engineered Systems (PISCES) models that encode industrially or societally relevant inputs and decode individually relevant outputs.
The PISCES Center takes an engineered-systems approach to precision modeling (Fig. 1). The foundations for knowledge, technology, and integration equally emphasize computation and experiment with the view that the two must cross-fertilize each other concurrently throughout the Center’s strategic plan. The premise of PISCES is that precision models combining population-scale variation with molecular–cellular biology will enable the simplest multicellular experimental system to give meaningful predictions. The PISCES Center will deliver simplified representations—models—of iPSC–organoid variation that are trained on individual exemplars and bootstrapped by computation to capture the population at large. The PISCES testbeds exploit Center expertise in gut, microbiome, and heart (Fig. 1), where sample availability is established and inter-individual differences are well documented. Ultimately, testbed-vetted PISCES models will create industry-grade products for patient stratification and direct-to-consumer products for genome contextualization. The PISCES Center will pursue its long-term vision through the following Research Thrusts:
We assert that an engineered-systems approach is uniquely powerful for pairing theory and experiment toward a predictive understanding of multicellular phenomena related to human populations. The discovery of new fundamental principles will feed directly into our overarching education, outreach, and workforce-development goal—to expand and diversify the pool of early-career engineers who are truly bilingual in the experimental and computational arms of precision modeling.