Machine Learning may help Predict when Immunotherapy is Effective

The immune system is a unique, complex, and pervasive system designed to recognizing problems, communicating with other cells, and performing their defensive functions to protect life. When it comes to defense, our bodies rely on attack thanks to the lymphatic and immune systems.

The immune system functions like the body’s own personal police force as it hunts down and eliminates pathogenic villains.

“The body’s immune system is very good at identifying cells that are acting strangely. These include cells that could develop into tumors or cancer in the future,” says Federica Eduati, Ph.D., assistant professor from the Department of Biomedical Engineering at Eindhoven University of Technology in Eindhoven, The Netherlands.

“Once detected, the immune system strikes and kills the cells,” Eduati added.

Stopping the attack
But it’s not always so straightforward as tumor cells can develop ways to hide from the immune system.

“Unfortunately, tumor cells can block the natural immune response. Proteins on the surface of a tumor cell can turn off the immune cells and effectively put them in sleep mode,” says Oscar Lapuente-Santana, Ph.D. researcher in the Computational Biology group at the Department of Biomedical Engineering at Eindhoven University of Technology in Eindhoven, The Netherlands.

Fortunately, there is a way to wake up the immune cells and restore their antitumor immunity, and it’s based on immunotherapy.

In the past few years, immunotherapy has revolutionized cancer treatment, especially based on antibodies targeting immune checkpoints, such as the cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein (PD-1), or its ligand (PD-L1).

Introducing immunotherapy
Immunotherapy is a cancer treatment that assists the immune system in its fight against cancer cells. One type of immunotherapy involves immune checkpoint blockers, which are drugs that tell the immune cells to ignore the shutdown orders coming from cancer cells.

The discovery of immune checkpoint blockers has been revolutionary for cancer treatment, with James P. Allison, Ph.D., professor, and chair of the Department of Immunology and the executive director of the immunotherapy platform at The University of Texas MD Anderson Cancer Center, where he also holds the Vivian Smith Distinguished Chair in Immunology and Tasuku Honjo, MD, Ph.D., professor emeritus at Kyoto University and distinguished professor and deputy director-general at the Kyoto University Institute for Advanced Study, jointly awarded the 2018 Nobel Prize in Physiology or Medicine for their work on immune checkpoint blockers.

Although immune checkpoint blockers have been successfully used to treat plenty of patients and different cancer types, only one-third of patients respond to the treatment.

“immune checkpoint blockers have had a big impact, but it could be bigger if we could figure out quickly which patients are most likely to respond to the treatment,” Eduati explained. “And it would also be great if we could understand why other patients are not responding to immune checkpoint blockers.”

To solve this problem, Lapuente-Santana, and Eduati, along with Maisa van Genderen, Ph.D. Candidate at Amsterdam University Medical Center in Amsterdam, The Netherlands, Peter Hilbers, Ph.D., full professor of BioModeling and BioInformatics at the department of Biomedical Engineering at Eindhoven University of Technology and Francesca Finotello, Ph.D., a PostDoc at the Biocenter – Division of Bioinformatics of the Medical University of Innsbruck (Austria), turned to machine learning to predict how patients might respond to ICB. Their work has just been published in the journal Patterns.

Searching the tumor microenvironment
To predict whether a patient will respond to immune checkpoint blockers, the researchers first needed to find particular biomarkers in tumor samples from the patients.

“Tumors contain more than just tumor cells, they also contain several different types of immune cells and fibroblasts, which can have a pro- or anti-tumor role, and they communicate with each other,” Lapuente-Santana explained.

“We needed to find out how complex regulatory mechanisms in the tumor microenvironment affect response to ICB. We turned to RNA-sequencing datasets to provide a high-level representation of several aspects of the tumor microenvironment,” Lapuente-Santana added.

To find the right mechanisms that could serve as biomarkers to predict patients’ response to immune checkpoint blockers, the team searched the microenvironment of tumors using computational algorithms and datasets from previous clinical patient care.

“RNA-sequencing datasets are publicly available, but the information about which patients responded to immune checkpoint blockers therapy is only available for a small subset of patients and cancer types,” Eduati noted. “So, we used a trick to solve the data problem.”

The trick
For their trick, instead of looking for the actual biological response to immune checkpoint blockers treatment, the researchers picked out several substitute immune responses from the same datasets. Despite not being the primary response to immune checkpoint blockers, together they could be used as an indicator of the effectiveness of immune checkpoint blockers.

Thanks to this approach, the team could use a large public dataset with thousands of patient samples to robustly train machine learning models.

“A significant challenge with this work was the proper training of the machine learning models. By looking at substitute immune responses during the training process, we were able to solve this,” Lapuente-Santana explained.

With the machine learning models in place, the researchers then tested the accuracy of the model on different datasets where the actual response to immune checkpoint blockers treatment was known. “We found that overall, our machine learning model outperforms biomarkers currently used in clinical settings to assess immune checkpoint blockers treatments,” says Eduati.

But why are Eduati, Lapuente-Santana, and their colleagues turning to mathematical models to solve a medical treatment problem? Will this replace the doctor?

“Mathematical models can provide a big picture of how individual molecules and cells are interconnected, while at the same time approximate the behavior of tumors in a particular patient. In clinical settings, this means that immunotherapy treatment with ICB can be personalized to a patient. It’s important to remember that the models can help doctors with their decisions on the best treatment, they won’t replace them.” Eduati said.

In addition, the model also helps in understanding which biological mechanisms are important for the biological response. Understanding and identifying the mechanisms that mediate immune checkpoint blockers response can guide how best to combine immune checkpoint blockers with other treatments to improve its clinical efficacy. However, this will first require experimental validation of the identified mechanisms before translating these results to clinical settings.

Dare to DREAM
The machine learning approach presented in the paper was also used by some of the researchers to take part in a DREAM challenge called “Anti-PD1 Response Prediction DREAM Challenge”.

DREAM is an organization dedicated to running crowd-sourced challenges involving algorithms in biomedicine. “We came first in one of the sub-challenges and competed under the name cSysImmunoOnco team,” adds Eduati.

Our immune system might be an efficient detective and disease hunter, but every now and then it needs a helping hand to eradicate elusive villains like cancer cells. Immunotherapy using immune checkpoint blockers is one such approach, but it doesn’t work for everyone.

Lapuente-Santana, Eduati, and colleagues have certainly dared to dream, and their work could prove pivotal in quickly identifying those who could be successfully treated with ICB in the future.

Thanks to machine learning, the researchers hope to rapidly deliver proper and effective cancer treatments to specific patients.

And for some cancer cells, it means that there could be no place to run, and no place to hide.

Reference
[1] Lapuente-Santana Ó, Van Genderen M, Hilbers P, Finotello F, Eduati F. Interpretable systems biomarkers predict response to immune checkpoint inhibitors Patterns, (2021) DOI:10.1016/j.patter.2021.100293

Featured image: Photo by Mathew Schwartz on Unsplash Used with permission.

Advertisement #5