June 25, 2024

MIT/Harvard Researchers Develop Computational Approach for Cellular Reprogramming

Researchers from MIT and Harvard University have developed a new computational approach that could significantly advance cellular reprogramming for applications in immunotherapy, cancer treatment, and regenerative medicine. Cellular reprogramming involves engineering a cell into a new state through targeted genetic interventions. By reprogramming a patient’s T-cells, for example, scientists aim to enhance their ability to kill cancer cells. However, due to the complexity of the human body’s 20,000 genes and over 1,000 transcription factors, finding the ideal genetic perturbation for a specific application has been challenging and costly.

The new computational approach developed by the researchers aims to efficiently identify optimal genetic perturbations using a smaller number of experiments compared to traditional methods. Their algorithm leverages the cause-and-effect relationship between factors in a complex system, such as genome regulation, to prioritize the best intervention in each round of sequential experiments. Through theoretical analysis and the application of their algorithms to real biological data, the researchers confirmed that their technique accurately identifies optimal interventions.

Typically, large-scale experiments are designed empirically, leading to high experimental costs. The researchers believe that their computational approach could reduce the number of trials needed to identify optimal interventions, thereby reducing experimental costs. The co-senior author of the study, Caroline Uhler, explains that a careful causal framework for sequential experimentation allows for the identification of optimal interventions with fewer trials.

When scientists design interventions for complex systems like cellular reprogramming, they often perform experiments sequentially. This process is ideally suited for active learning, a machine-learning approach. Active learning involves collecting data samples and using them to learn a model of the system, incorporating the knowledge gathered so far. The model is then used to design an acquisition function that evaluates potential interventions and selects the best one to test in the next trial.

However, traditional acquisition functions based on correlation between factors are not effective for complex problems, resulting in slow convergence. The MIT and Harvard researchers addressed this limitation by focusing on the causal structure of the system. Their algorithm ensures that the models of the system account for causal relationships and the acquisition function evaluates interventions based on these causal relationships. This approach allows for the prioritization of informative interventions that are most likely to lead to the optimal intervention in subsequent experiments.

To improve the efficiency of their acquisition function, the researchers used a technique called output weighting, inspired by the study of extreme events in complex systems. By emphasizing interventions that are likely to be closer to the optimal intervention, they were able to further enhance the efficiency of their approach.

The researchers tested their algorithms using real biological data in a simulated cellular reprogramming experiment. The acquisition functions consistently identified better interventions than baseline methods throughout the multi-stage experiment. Even if the experiment was terminated at any stage, their approach still proved to be more efficient than baseline methods.

The researchers are currently collaborating with experimentalists to apply their technique to cellular reprogramming experiments in the laboratory. They believe that their approach could also be applied to problems outside of genomics, such as identifying optimal prices for consumer products or enabling optimal feedback control in fluid mechanics applications.

In the future, the researchers plan to enhance their technique to optimize beyond desired mean shifts and explore the use of artificial intelligence to learn causal relationships in the system. Overall, their computational approach for cellular reprogramming shows great potential in advancing the field of immunotherapy, cancer treatment, and regenerative medicine.

 

*Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it