Advancements in diagnostics are shaping the future of personalized medicine, particularly in the fight against complex, hard-to-cure diseases like cancer and diabetes. At the forefront of this innovation are computational models and deep learning technologies that are transforming how we predict, diagnose, and treat diseases.
By combining cutting-edge biological research with engineering and computational sciences, researchers at Northeastern University’s Roux Institute are paving the way for more precise and effective treatments. One of the key innovators in this space is Dr. Kiran Vanaja, Assistant Research Professor at Northeastern’s Roux Institute and the Bioengineering Department. His pioneering work brings revolutionary approaches to cancer diagnostics and opens new possibilities for managing diseases across multiple fields.
An Innovative Approach to Cancer Diagnostics
Cancer remains a formidable opponent in medicine due to its complex and resilient nature. Billions of dollars are spent on cancer research each year in the US alone.
One of the biggest challenges is that cancer cells are highly adaptable, making single-drug treatments ineffective over time as the cells develop drug resistance.
That is why the Vanaja Lab takes a different approach. Drawing from his background in electrical engineering, biomedical engineering, and systems biology, Dr. Vanaja’s lab focuses on constructing computational and mathematical models that predict how cancer cells will adapt to treatments. Rather than attempting to target a single mutation, their goal is to outsmart cancer cells with multi-step, strategic therapeutic interventions.
“Any simple approach of blocking a mutation in cancer is unlikely to work because the cancer cell will keep adapting, as they have for billions of years,” says Vanaja.
Therefore, the research centers on understanding how cancer cells behave when they survive initial treatments. Dr. Vanaja explains it this way: if you put cancer cells in a dish and treat them with an FDA-approved drug, approximately 99% of all the cells will be killed. However, there will be small pockets of cells that survive. When you look at this at scale, however, such as in a sample of 10 million cancer cells, that means 100,000 cancer cells survive.
By all conceivable measurable properties (cell phenotypes), these remaining cells are completely different from the cells that started out. Their migration, metabolic status, reactive oxygen species, epithelium, and mesenchymal nature are all changed in every possible manner.
“We wanted to determine how to characterize a cell like this,” said Vanaja. “Because what we observe is the remaining tumor cells are entirely changed and become almost like a demonic form of what they were, having acquired all these hard-to-treat properties that we don’t want them to have.”
A Multi-Step Strategy to Outsmart Cancer
Dr. Vanaja’s research aims to develop strategies that move cancer cells into vulnerable positions where they can be better targeted by subsequent treatments. By using deep learning models, his team is creating multi-step therapeutic interventions designed to move cancer cells through a series of treatment stages, increasing the chances of eliminating them 100%.
“If you can move these cells to a space where we know a drug will work, you increase the chances of killing them,” Vanaja says. “It’s not about just blocking one pathway that’s mutated, but guiding the cells to a spot where they can be further targeted.”
For example, a tumor sample from a patient could be treated with several drugs in a specific order. Drug A might be applied first, then Drug B, and then Drug C, guiding the cancer cells into a state where they are more likely to die. This approach, developed over nearly a decade of research, leverages recent advancements in DNA, RNA, and proteomics sequencing as well as deep learning technologies to map cellular behaviors and predict drug responses.
Vanaja explains: “We now can measure almost everything inside the cell. Deep learning has completely blown open the entire field of biology.”
Commercialization Challenges and the Need for Industry Collaboration
There is much work to be done before these breakthroughs can be applied broadly in clinical settings. Dr. Vanaja’s approach relies on using deep learning neural networks and vast datasets to map cellular responses to treatments. Creating these models requires not only data on a wide variety of cancer cell types but also extensive testing in clinical trials.
One of the challenges is the sheer complexity of the task. Cancer affects nearly every tissue type in the body, and each type behaves differently. For example, the RNA sequencing of prostate cells will have different interconnections compared to breast or pancreatic cells. Developing a universal mapping of cellular responses requires data from hundreds of different cell types, which is both time-consuming and expensive.
“There’s a technique called site-seq, which allows us to simultaneously read the inside and outside of a cell, giving us a proxy for mapping cellular phenotypes,” Vanaja explains. “But if we want to collect data from 100 different cell types, it could cost close to a million dollars just for the assays.”
Building these models requires collaboration across academia and industry, with funding being a significant barrier. Dr. Vanaja envisions that it will take 5 to 10 years to bring these models to fruition, but with partnerships and support, these diagnostic tools could revolutionize the way we treat cancer and other chronic diseases.
Looking Forward: The Future of Cancer Diagnostics and Beyond
For the Vanaja Lab, the long-term goal is not to find a single drug that cures cancer but to use a combination of existing therapies in a new, more strategic way. By understanding how cells behave when they are inhibited, and by leveraging computational models, they strive to devise more effective ways to target and kill cancer cells.
“The idea behind our approach is not to come up with newer drugs, as researchers have spent the last several decades developing what seems like every possible single-molecule drug,” says Dr. Vanaja. “So our goal is not to find the silver bullet, but rather to find a combination of drugs that can cure these historically challenging-to-treat diseases.”
Ultimately, this approach aims to transform cancer from a deadly, unpredictable disease into a manageable condition. With further research and collaboration, deep learning models like those developed by Dr. Vanaja and his team could play a crucial role in the future of diagnostics and disease management.
Learn more about Dr. Vanaja and his research here.
Interested in this technology? Contact Senior Commercialization Manager, Vaibhav Saini.
Written by Elizabeth Creason