Memorial Sloan Kettering develops AI cancer treatment predictor

A new artificial intelligence based model called SCORPIO has been developed by researchers from Memorial Sloan Kettering Cancer Center and the Tisch Cancer Institute at Mount Sinai to predict whether cancer patients will benefit from immune checkpoint inhibitors, a type of immunotherapy. The model uses routine blood tests and clinical data to make predictions, making it more accessible and cheaper than current methods.

According to Dr Luc Morris, a surgeon and research lab director at MSK, the model outperforms the tests currently used to predict outcomes. The study, which included nearly 10,000 patients across 21 different cancer types, was led by Dr Morris and Diego Chowell, an Assistant Professor of Immunology and Immunotherapy at Mount Sinai. The research was supported by the National Institutes of Health and other organizations and has the potential to make immunotherapy more widely available and effective for cancer patients.

Introduction to Immunotherapy and Checkpoint Inhibitors

Immunotherapy has revolutionized the field of oncology, offering new hope for cancer patients worldwide. Among the various immunotherapeutic approaches, checkpoint inhibitors have emerged as a powerful tool in the fight against cancer. These drugs work by releasing the brakes on immune cells, allowing them to recognize and attack cancer cells more effectively. However, despite their potential, checkpoint inhibitors do not benefit all patients, and their use is often limited by the lack of reliable biomarkers to predict treatment response.

The current standard for predicting response to checkpoint inhibitors relies on two FDA-approved biomarkers: tumor mutational burden (TMB) and PD-L1 immunohistochemistry. TMB measures the number of mutations in a tumor, while PD-L1 immunohistochemistry evaluates the expression of the programmed death-ligand 1 protein in tumor samples. However, both of these methods require tumor samples to be collected, which can be invasive and may not always be feasible. Furthermore, genomic testing to assess TMB is expensive and not widely available, and there is significant variability in evaluating PD-L1 expression.

Development of the SCORPIO Model

To address these limitations, researchers have developed a new model called SCORPIO, which uses readily available clinical data, including routine blood tests performed in clinics around the world. The model relies on ensemble machine learning, a type of artificial intelligence that combines several tools to look for patterns in clinical data from blood tests and treatment outcomes. The development of SCORPIO involved collecting data from over 2,000 patients at Memorial Sloan Kettering (MSK) who had been treated with checkpoint inhibitors, representing 17 different types of cancer.

The model was then tested using data from an additional 2,100 MSK patients to verify its accuracy in predicting treatment outcomes. Further validation was performed using data from nearly 4,500 patients treated with checkpoint inhibitors in 10 different phase 3 clinical trials from around the world, as well as from nearly 1,200 patients treated at Mount Sinai. The study includes nearly 10,000 patients across 21 different cancer types, representing the largest dataset in cancer immunotherapy to date.

Performance of the SCORPIO Model

The results of the study demonstrate that the SCORPIO model outperforms the currently used tests in the clinic, including TMB and PD-L1 immunohistochemistry. The model’s simplicity and affordability make it an attractive option for ensuring more equitable access to care, reducing costs, and helping patients receive treatments that are most likely to benefit them individually.

The SCORPIO model uses a combination of clinical variables, including complete blood count and comprehensive metabolic profile, which are readily available in clinics worldwide. This approach eliminates the need for invasive tumor sampling and expensive genomic testing, making it a more accessible and cost-effective option for patients and healthcare providers.

Future Directions and Clinical Implications

The next steps for the SCORPIO model involve collaborating with hospitals and cancer centers around the world to test the model with additional data from a wider variety of clinical settings. The feedback received will help optimize the model, ensuring its widespread applicability to patients and physicians in different locations. Additionally, work is underway to develop an interface that is readily accessible by clinicians, regardless of their location.

The clinical implications of the SCORPIO model are significant, as it has the potential to revolutionize the way checkpoint inhibitors are used in cancer treatment. By providing a more accurate and reliable method for predicting treatment response, the model can help ensure that patients receive the most effective treatment for their individual needs, reducing the risk of unnecessary side effects and improving overall outcomes.

Limitations and Future Research Directions

While the SCORPIO model shows great promise, there are several limitations that need to be addressed in future research. One of the main limitations is the potential for bias in the dataset used to develop the model. The study included a large number of patients from MSK, which may not be representative of the broader population. Further validation studies are needed to confirm the model’s performance in different clinical settings and patient populations.

Another area of future research involves exploring the underlying biological mechanisms that drive the predictive power of the SCORPIO model. By understanding how the model works, researchers can identify new biomarkers and develop more effective treatments for cancer patients. Additionally, the integration of the SCORPIO model with other emerging technologies, such as liquid biopsies and artificial intelligence-powered diagnostic tools, may further enhance its clinical utility and improve patient outcomes.

Ethical Considerations and Disclosures

The study was supported by several funding agencies, including the National Institutes of Health, the Department of Defense, and various foundations. Several authors have filed a provisional patent application for using routine blood tests and clinical variables to predict cancer immunotherapy response. Others are co-inventors on patents related to tumor mutational burden and multimodal machine learning models for predicting immunotherapy response. Some authors also report consulting work for pharmaceutical companies unrelated to the current research. These disclosures highlight the need for transparency and rigor in scientific research, ensuring that conflicts of interest do not compromise the integrity of the study or its findings.