People often compare the genome to a computer’s program, with the cell using its genetic code to process environmental inputs and produce appropriate responses.
But the machine metaphor can be extended even further to any biological system, and applying established concepts of engineering to biology could revolutionize how scientists make their observations within biology, according to research from University of Michigan.
In a paper published in PNAS, Indika Rajapakse, Ph.D., Joshua Pickard, Ph.D. (now an Eric and Wendy Schmidt Postdoctoral Fellow at the Broad Institute), and their team propose that fundamental principles of control theory and observability can be applied to study biological processes that change over time.
Control theory and observability were pioneered in the 1960s by Elmer Gibert, Ph.D., an engineer who spent most of his career at U-M.
A chance encounter at a lecture between Gilbert and Rajapakse sowed the seeds for the melding of math, engineering and biology.
“Control theory, or controllability, essentially means how to steer a system to something else and what inputs you need to give to a system to steer it in that direction,” explained Rajapakse.
For example, a cell undergoes differentiation, becoming specialized, based on its exposure to a transcription factor–such as the 2012 Nobel Prize winning discovery that a skin cell can be reprogrammed into a stem cell, a discovery that has revolutionized biomedical research.
The related engineering concept of observability refers to the minimum number of signals that will indicate the status of a system.
When applied to biology, report the Rajapakse team, observability can help researchers discover biomarkers.
“Dynamics is one of the most important concepts in all of biology,” Rajapakse said.
“You can measure status at one point in time, but biological systems change over time.”
Observability theory provides a generalist framework for selecting biomarkers.
“Most existing biomarker discovery methods are limited to a single type of data,” Pickard said.
“Our approach, by contrast, works across different data and experimental systems. In the paper, we show this by applying our algorithms to transcriptomics, chromatin structure, and neural imaging and EEG datasets.”
As an example, the team applied their approach to multiple time-series transcriptomics datasets, including studies of cell reprogramming, pesticide exposure, and the cell cycle.
Through Dynamic Sensor Selection, they pinpointed biomarkers at each time point and showed that these reduced data were enough to represent the full behavior of the system.
“The idea is identifying the minimal number of variables where if I monitor those, I can say something about the whole system,” said Rajapakse.
“Studying the entire genome is extremely expensive and time consuming. DDS provides a way to study a subset of data and then from that, I can reconstruct the entire genome.”
The findings hold implications for development of more efficient experiments and even early detection of disease.
“The concept of observability will guide us to detect whether what is expected is happening or not. If not, I can abort the experiment and save money.
Likewise, I can identify a biomarker wherein if it is changing in a certain way that is associated with a disease, like cancer, I can perhaps prevent the disease from occurring,” he said.
Additional authors: Cooper Stansbury, Amit Surana, Lindsey Muir, Anthony Bloch
Funding/disclosures: This work is supported by Air Force Office of Scientific Research (AFOSR) under award number FA9550-22-1-0215 (IR,AB), FA9550-23-1-0400 (AB), NSF DMS-2103026 (AB) and NIGMS GM150581 (JP).
Paper cited: “Dynamic Sensor Selection for Biomarker Discovery,” PNAS. DOI: 10.1073/pnas.2501324122
Journal
Proceedings of the National Academy of Sciences
Article Title
Dynamic Sensor Selection for Biomarker Discovery