Biological engineers have created a computational method to extract valuable insights from extensive biological datasets. Their research successfully decoded the interactions that dictate how the immune system reacts to tuberculosis vaccinations and infections.
In the previous two decades, advancements in technology have enabled scientists to collect an immense amount of biological data. Large-scale studies in areas like genomics, transcriptomics, proteomics, and cytometry can yield vast data volumes from cellular or multicellular systems.
However, interpreting this data can be challenging, particularly when analyzing intricate systems like the interaction sequences that take place when the immune system confronts a foreign pathogen.
MIT biological engineers have now introduced a new computational strategy to derive insightful information from these datasets. By employing this novel technique, they were able to untangle a series of interactions that dictate how the immune system reacts to tuberculosis vaccination and infection.
This approach could prove beneficial for vaccine developers as well as researchers investigating complex biological systems, according to Douglas Lauffenburger, the Ford Professor of Engineering in the departments of Biological Engineering, Biology, and Chemical Engineering.
“We’ve developed a computational modeling system that predicts the impact of variations in extremely complex systems, involving multiple scales and diverse components,” says Lauffenburger, the primary author of this fresh study.
Shu Wang, a former postdoctoral researcher at MIT who is now an assistant professor at the University of Toronto, and Amy Myers, a research manager in the lab of University of Pittsburgh School of Medicine Professor JoAnne Flynn, are the lead authors of a recently published paper in the journal Cell Systems.
Understanding complex systems
When exploring complex biological systems such as the immune system, scientists can gather various types of data. Genome sequencing reveals which gene variants a cell holds, while examining messenger RNA transcripts shows which genes are being expressed in that cell. Proteomics allows researchers to analyze the proteins present in a cell or biological system, and cytometry quantifies the multitude of cell types involved.
Utilizing computational methods like machine learning, scientists can harness this data to craft models that predict a specific outcome based on particular inputs — such as whether a vaccine will trigger a substantial immune response. Nonetheless, that form of modeling fails to highlight the intermediary steps connecting the input to the output.
“That AI approach can indeed have significant clinical utility, yet it doesn’t effectively illuminate biological processes since the focus is on what transpires between the inputs and outputs,” Lauffenburger notes. “What mechanisms actually translate inputs into outputs?”
To form models that reveal the intricate processes within complex biological systems, the researchers utilized models known as probabilistic graphical networks. These models depict each measured variable as a node, creating maps of how each node interconnects with others.
Probabilistic graphical networks are frequently employed in fields like speech recognition and computer vision, but their application in biology has been limited.
Lauffenburger’s lab previously leveraged this modeling technique to examine intracellular signaling pathways, which entailed analyzing a singular type of data. To expand this methodology to encompass various datasets simultaneously, the researchers employed a mathematical technique capable of filtering out correlations between variables that do not directly influence one another. This method, known as graphical lasso, is a variation of techniques generally used in machine learning to eliminate results likely caused by noise.
“In correlation-based network models generally, a common challenge is that everything appears to affect everything else, making it necessary to identify the most critical interactions,” Lauffenburger says. “Using probabilistic graphical network frameworks enables one to hone in on the most probable direct interactions and discard those likely being indirect.”
Understanding the vaccination mechanism
To validate their modeling approach, the researchers analyzed data from studies of a tuberculosis vaccine, known as BCG, which is an attenuated form of Mycobacterium bovis. This vaccine is administered in various countries where TB is prevalent, though its effectiveness can be inconsistent, and its protective effects may diminish over time.
To enhance TB protection, researchers have been exploring whether administering the BCG vaccine via intravenous injection or inhalation could elicit a more effective immune response compared to traditional injection methods. Animal studies indicated that intravenous delivery significantly improved the vaccine’s effectiveness. In the MIT study, Lauffenburger and his team sought to uncover the underlying mechanisms responsible for this enhancement.
The data analyzed encompassed around 200 variables, including levels of cytokines, antibodies, and various immune cell types collected from about 30 animals.
Measurements were taken before vaccination, following vaccination, and after TB infection. Through their innovative modeling approach, the MIT team identified the processes necessary to generate a robust immune response. Their findings indicated that the vaccine activates a particular subset of T cells that produce cytokines, stimulating B cells to create antibodies that specifically target the bacterium.
“Almost like a roadmap or subway map, we were able to pinpoint the most significant pathways. Despite numerous other changes occurring within the immune system, they were largely irrelevant to the critical processes,” Lauffenburger remarks.
The researchers subsequently utilized the model to predict the impacts of specific disruptions, such as the suppression of certain immune cell subsets, on the system. The model forecasted that nearly eliminating B cells would minimally affect the vaccine response, which experimental results corroborated.
This modeling technique could assist vaccine developers in forecasting the effects of their vaccines and making necessary adjustments to improve their efficacy prior to human trials. Lauffenburger’s lab is currently applying this model to study the mechanisms of a malaria vaccine administered to children in Kenya, Ghana, and Malawi over recent years.
Additionally, the lab is employing this modeling approach to investigate the tumor microenvironment, comprising various immune and cancer cells, aiming to predict how tumors will react to different treatment modalities.
