In the last twenty years, advancements in technology have empowered scientists to collect a remarkable amount of biological data. High-throughput experiments in genomics, transcriptomics, proteomics, and cytometry yield vast datasets from various cellular or multicellular systems.
Yet, deciphering this complex information can be challenging. This complexity is especially pronounced when studying intricate interactions, such as those observed when the immune system engages with foreign pathogens.
Recently, biological engineers at MIT have pioneered a novel computational method to extract valuable insights from these extensive datasets. Utilizing their innovative technique, researchers have successfully illuminated the interactions that dictate how the immune system reacts to tuberculosis vaccination and subsequent infections.
Douglas Lauffenburger, the Ford Professor of Engineering in the Biological Engineering, Biology, and Chemical Engineering departments, notes that this approach could prove beneficial for both vaccine developers and researchers investigating various complex biological systems.
“We’ve developed a computational modeling framework capable of predicting effects of perturbations within highly complex systems, incorporating multiple scales and diverse components,” says Lauffenburger, who is the senior author of the study.
Shu Wang, a former postdoc at MIT now serving as an assistant professor at the University of Toronto, alongside Amy Myers, a research manager in the lab of JoAnne Flynn at the University of Pittsburgh School of Medicine, are the lead authors of the research, which was published today in the journal Cell Systems.
Unraveling Complex Biological Systems
When delving into complex biological systems like the immune system, scientists can gather a myriad of data types. By sequencing cell genomes, they ascertain gene variants, while examining messenger RNA transcripts reveals active gene expressions. Proteomics allows researchers to analyze protein levels within a cell or biological system, and cytometry quantifies various cell types present.
Employing computational methods, particularly machine learning, scientists can train models to forecast outcomes based on particular inputs, such as predicting the robustness of a vaccine-induced immune response. However, these models often provide limited insights regarding the intermediary processes that connect inputs to outputs.
“While the AI approach is valuable in clinical settings, it falls short in elucidating biological intricacies, as it overlooks critical changes happening between the input and output,” Lauffenburger explains. “We need to understand the mechanisms that produce outputs from these inputs.”
To build models capable of pinpointing the underlying mechanisms within complex biological systems, the researchers employed probabilistic graphical networks. These models represent each variable as a node, mapping the connections among them.
Probabilistic graphical networks are frequently utilized in fields like speech recognition and computer vision, yet their application within biology remains limited.
Lauffenburger’s lab has previously harnessed these models for analyzing intracellular signaling pathways, which involved interpreting a singular type of data. To modify this method for analyzing multiple datasets simultaneously, the researchers introduced a mathematical technique capable of filtering out spurious correlations between variables. Known as graphical lasso, this technique is a refinement often used in machine learning to eliminate results likely influenced by noise.
“Correlation-based network models often suggest that all variables influence one another, complicating the identification of essential interactions,” Lauffenburger points out. “Employing probabilistic graphical network frameworks enables us to focus on the most direct interactions while discarding less relevant correlations.”
Decoding the Mechanism of Vaccination
The researchers tested their modeling approach using data from studies on a tuberculosis vaccine known as BCG, derived from a weakened version of Mycobacterium bovis. BCG is commonly utilized in countries where tuberculosis is prevalent, although it doesn’t always provide effective immunity, and its efficacy may diminish over time.
In their quest for enhanced tuberculosis protection, researchers have examined whether intravenous or inhaled delivery of the BCG vaccine elicits a stronger immune response compared to traditional injections. These investigations, conducted in animal models, revealed that intravenous administration significantly improved vaccine efficacy. Lauffenburger and his team aimed to uncover the underlying mechanism driving this success.
The data analyzed in this study comprised approximately 200 variables, including cytokines, antibodies, and various immune cell types, collected from around 30 animals pre-vaccination, post-vaccination, and after TB infection.
By employing their innovative modeling technique, the MIT team discerned the essential steps for generating a robust immune response. They found that the vaccine activates a specific subset of T cells, which, in turn, produces a cytokine triggering B cells to generate antibodies targeting the bacterium.
“It’s akin to following a roadmap; we identified the most crucial paths. Although numerous changes occur within the immune system, many are peripheral and less significant,” Lauffenburger shares.
The researchers further utilized the model to predict the potential impact of specific disruptions on the system, such as the suppression of immune cell subsets. The model suggested that significantly reducing B cells would have minimal effect on the vaccine’s efficacy, a prediction that subsequent experiments confirmed.
This modeling strategy presents a valuable tool for vaccine developers, enabling them to project the potential effects of their vaccines and refine their formulations prior to human trials. Currently, Lauffenburger’s lab is leveraging this model to explore the mechanisms of a malaria vaccine administered to children in Kenya, Ghana, and Malawi over the past several years.
Additionally, the lab applies this modeling technique to investigate the tumor microenvironment, which encompasses various immune and cancerous cell types, aiming to anticipate how tumors might respond to different therapies.
This groundbreaking research received funding from the National Institute of Allergy and Infectious Diseases.
Photo credit & article inspired by: Massachusetts Institute of Technology
