The development of checkpoint inhibitor therapies is undoubtedly a major advance in cancer treatment. But first, these drugs must undergo extensive clinical research, which requires precision, efficacy, and speed before they reach the doctor’s office. It’s time for a broader approach – artificial intelligence powered biomarker analysis for immune profiling.
Why does lifesaving cancer treatment discriminate?
The field of immunotherapy has brought new breakthroughs in cancer treatment as well as new hope for patients and their families. For example, just recently, another checkpoint inhibitor drug, pembrolizumab (KEYTRUDA, Merck), was FDA approved for patients with PD-L1 expressing tumors of squamous cell carcinoma of the esophagus (ESCC).
These targeted drugs are intended to interact with checkpoint inhibitors to release the power of the immune system to recognize the body’s mutated cancer cells. And yet, this breakthrough treatment simply doesn’t work on everyone. In fact, the biggest hurdle is understanding why patients don’t respond to therapy or relapse later on.
We need better ways to stratify patients.
Precision immuno-oncology requires understanding of the tumor micro-environment to accurately gauge tumor heterogeneity and immune contexture. And currently, checkpoint inhibitors work best on advanced mutated cancer cells in cases such as melanoma, kidney, bladders, and (NSCLC) lung cancer.
One approach is to better understand the patient’s genetic biomarkers so we can predict his or her immune response, such as micro-satellite instability (MSI) or mismatch repair (dMMR).
The search for predictive biomarker signatures
One of the key pieces to understanding each patient’s response to cancer treatment is to understand the unique landscape of cell populations within the tumor micro-environment. But there is more to learn beyond the tumor-host interactions revealed by digital pathology. By combining this data with additional health information about the patient and comparing it to other patient cohorts, we can get a fuller picture of what makes some patients successfully respond to new therapies and others not.
AI gives us the power to predict patient response
Microscopy-based pathology can give us incredible insight into predicting the efficacy of a cancer drug in attempts to explain common hurdles such as drug resistance and heterogeneity of tumors. But manual scoring and the human eye have limitations, especially when it comes to comparing vast quantities of patient data and images. This is where artificial intelligence becomes a pathologist's ally when machine learning algorithms can be used to find specific patterns and make sense of the data.
Using AI, Tissue Phenomics takes it a step further by analyzing tissue images to reveal complex interactions in the tumor micro-environment. This makes it finally possible to predict response using biomarker signatures to accurately identify drug responders. We find these by evaluating multiplex biomarker assays integrated with multi-omics data to characterize the involved populations of the tumor micro-environment and thus gain a better understanding of the individual tumor-host situation of each single patient.