Computational tool helps uncover gene networks of cell fate
Multi-cellular organisms are made up of a multitude of cell types and much of the information of what the “type” a cell is, is defined by the set of genes the cell activates. Gene regulatory networks (GRNs) are molecular networks operating within each cell that determines which genes must be expressed when and where, thereby conferring the type of a cell. Mistakes in the expression of genes can make cells go awry, losing their identity and often reaching undesired states such as cancer. However, to date, we only had a coarse understanding of these gene regulatory networks that lacked the resolution of individual cell types and more importantly how they change between normal and disease cell types.
New research from the Wisconsin Institute for Discovery (WID) may provide key insights that will aid researchers to construct a more precise view of what drives cellular identity. The study published in Nature Communications describes a computational approach, scMTNI, that combines the power of single-cell omics datasets with machine learning tools to examine how and which gene regulatory networks set a cell on its path to reach a particular endpoint. Identification of gene regulatory networks that underlie the type and identity of cells is at the heart of understanding many disease and normal biological processes.
“We are trying to understand the causal mechanisms of how cells transition from one state, e.g., a pluripotent state to a more differentiated state, e.g. a skin or nerve cell and vice versa. We want to know how different normal cell types emerge and what might lead to cells becoming aberrant.” says Sushmita Roy, associate professor in the Department of Biostatistics and Medical Informatics and faculty member in the Wisconsin Institute for Discovery. Single cell genomics has revolutionized biomedicine by providing a high-dimensional measurement of individual cells. However, we need tailored computational tools that can make sense of these measurements to tell us what molecular networks are associated with cells exhibiting specific measurement profiles. To gain such an understanding “we are developing computational tools that can integrate large-scale molecular profiles measured for each individual cell in a population of thousands of cells to define these GRNs” says Roy.
Current methods for building GRNs do not accurately model population heterogeneity and the dynamics of cell state transitions. However, it’s important to take into account the structure of the cell population when defining the networks. This helps us understand how the networks change over time and how they differ between different types of cells.
To address this challenge, Dr Roy and her graduate student Shilu Zhang developed a framework called single-cell Multi-Task Network Inference (scMTNI). Utilizing both simulated and real datasets from different developmental contexts, they were able to identify key GRNs that are likely crucial for the dynamics of the system but were missed by other methods.
The Roy lab is at the forefront of cross-disciplinary partnerships bridging computational science with biology. One such collaboration involves fellow WID faculty Rupa Sridharan, who is researching an experimental process called cellular reprogramming. Cellular reprogramming is the process of turning unipotent cells of the body into pluripotent cells which have the ability to become any of the different cell types in an organism. Unfortunately, the process of reprogramming is inefficient. It is at this point that Roy hopes to help Dr. Sridharan and other researchers by providing a computational tool that can identify genes that when turned on or off could make reprogramming more efficient. The computational tool would trace the path or network of cells and monitor the changes of a cell through its lifespan. With a detailed map, researchers can then use this information to determine when and where the cells become differentiated.
Ultimately, Roy hopes the scMTNI tool will help researchers such as developmental biologists, stem cell engineers and clinicians to understand the molecular underpinnings of why some cells transition from normal to disease states, guide more efficient strategies to develop patient-specific disease models, as well as to better understand mechanisms of host-microbe interactions. “Identifying the GRN components that drive cell fate transitions is a very hard problem, and requires rich molecular profiling data, iterative model building and refinement through informative experiments” Roy says. “But with collaborative, interdisciplinary efforts we can achieve these goals and help advance our understanding of GRNs in cell fate and type specification.”
–By Laura C. RedEagle
Funding
We thank the Center for High Throughput Computing at University of Wisconsin-Madison for computational resources. This work is supported by the National Institutes of Health NIGMS grant 1R01GM117339, 1R01GM144708-01A1, 2R01GM113033 and the Department of Energy grant DE-SC0021052.