Search...
Explore the RawNews Network
Follow Us

Deciphering immune response mechanisms using probabilistic graphical networks.

[original_title]
0 Likes
November 6, 2024

Over the past 20 years, new technologies have allowed scientists to generate enormous quantities of biological data. Experiments spanning genomics, transcriptomics, proteomics and cytometry produce large volumes of information about cells or multicellular systems in which experiments take place.

As is often the case when dealing with complex systems like those associated with immunity response mechanisms such as those seen when facing off against foreign pathogens, making sense of this information can sometimes be tricky. This is particularly evident when looking into interactions that occur in response to exposure of foreign pathogens by immune cells.

MIT biological engineers have devised an innovative computational method for extracting useful information from these datasets. With it, they were able to uncover interactions that influence how an immune system reacts against tuberculosis vaccination and subsequent infection.

Douglas Lauffenburger, Ford Professor of Engineering for Biological, Cell and Chemical Engineering departments. says this approach could prove invaluable when creating vaccines or studying any complex biological system.

We have successfully developed a computational modeling framework which permits predictions of effects of perturbations to highly complex systems that span multiple scales and feature numerous components, while still remaining highly realistic in their representations of reality.
Douglas Lauffenburger is Ford Professor of Engineering from Massachusetts Institute of Technology’s departments of Biological, Electrical and Mechanical Engineering.

Shu Wang, an MIT postdoc who now teaches at the University of Toronto as an assistant professor, and Amy Myers – research manager of Professor JoAnne Flynn from Pitt School of Medicine’s lab – are lead authors in a new paper released today in Cell Systems journal detailing this work.

Modelling Complex Systems
When studying complex biological systems such as the immune system, scientists can collect many different kinds of data. Sequencing cell genomes provides them with information on which gene variants a cell carries while messenger RNA transcript analysis shows them which genes are expressed by that particular cell. Proteomics allows scientists to measure proteins present within biological systems or cells while Cytometry helps quantify all available cell types present.

Engineers use computational approaches like machine learning to use this information and train models that predict specific outputs based on inputs; for instance, whether or not vaccination will cause an effective immune response. Unfortunately, such modeling doesn’t reveal anything about what’s happening between input and output.

“Artificial Intelligence can be extremely valuable for clinical medical uses; however, its application to biological understanding has its limitations,” Lauffenburger notes. “You want to know about everything going on between inputs and outputs – what mechanisms actually create output from input?”

Researchers turned to probabilistic graphical networks as an approach for creating models capable of simulating complex biological systems, in order to understand their inner workings. These models represent each measured variable as nodes connected by lines on maps that show them all interrelate with each other.

Probabilistic graphical networks have long been employed for applications including speech recognition and computer vision; however, their applications in biology remain relatively underused.

Lauffenburger’s lab previously employed this approach to analyze intracellular signaling pathways, which required only looking at one type of data. To apply it to numerous datasets at once, researchers applied a mathematical technique known as “graphical lasso,” an adaptation of machine learning models used to filter out correlations that do not directly impact each other and remove results due to noise.

Lauffenburger points out the challenges with correlation-based network models are their tendency for interdependencies among interactions to influence one another; therefore, one must identify essential relationships. With probabilistic graphical network frameworks however, one can focus on those things likely to have direct connections while simultaneously eliminating indirect ones.

Mechanism of Immunisation
To test their modeling approach, researchers collected data from studies of tuberculosis vaccine BCG; which contains attenuated strains of Mycobacterium bovis. BCG vaccination is widely used but may become ineffective with time as its effectiveness wears away over time.

Researchers looking for more effective TB protection have conducted numerous animal experiments testing whether intravenously administered BCG vaccination could produce a more powerful immune response than injecting it directly. Animal studies concluded that intravenous administration proved much more successful; Lauffenburger and his colleagues at MIT explored its mechanism.

Researchers involved in this research conducted measurements on 200 variables relating to levels of cytokines, antibodies and specific immune cell types from about 30 animals.

Measurements were collected before, after, and post TB infection vaccination with their new modeling approach from MIT team were then able to analyze this data in order to establish steps needed for an effective immune response triggered by vaccination: specifically T cells producing cytokine which then activated B cells which produced antibodies targeting the bacteria.

“Lauffenburger says this analysis revealed the crucial pathways, even as other parts of the immune system changed in various ways; these changes, though significant, weren’t important compared with others that might happen along these critical routes and didn’t matter nearly as much,” Lauffenburger says.

Researchers then employed this model to make predictions regarding how specific disruptions, like suppressing certain immune cells, would impact their system. Experiments verified this prediction that when B cells were nearly eradicated there would be minimal impact on vaccine response; and predictions made using simulation software confirmed their accuracy.

Modeling approaches such as these could provide vaccine developers with an accurate way to predict the effects of their vaccines and make tweaks that would enhance them before being given to humans for testing. Lauffenburger’s lab has utilized their model extensively over recent years when investigating malaria vaccine distribution to children across Kenya, Ghana and Malawi.

His laboratory employs this modeling approach to investigate the tumor microenvironment, composed of various immune and cancerous cells, in an attempt to predict how tumors might respond to specific treatments.

This study was funded by the National Institute of Allergy and Infectious Diseases.
Source:Massachusetts Institute of Technologyjournal reference:Wang S et al (2024) A Markov Field Network Model with Multimodal Data Prediction to Explore Effects of Immune System Disruptions on Intravenous BCG Vaccination in Macaques Cell Systems: doi.org/10.1016/j.cels.2024.10.001

Social Share
Thank you!
Your submission has been sent.
Get Newsletter
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus

Notice: ob_end_flush(): Failed to send buffer of zlib output compression (0) in /home3/n489qlsr/public_html/wp-includes/functions.php on line 5464