Something happened to Jimmy Carter’s expected demise in 2015. The former US president was diagnosed with a previously untreatable, and particularly virulent, form of brain cancer.
By Halima Bensmail
Doha – Every newspaper in the world was diligently writing their obituaries and preparations were already underway for a state funeral. There was a collective sigh in America as it was painfully obvious the country was going to lose one of its most beloved former presidents.
Outside of a few oncologists, nobody knew about the breakthrough cancer drug Pembrolizumab, or, as it is now known, ‘the Jimmy Carter drug’. A combination of a fast diagnosis made possible by artificial intelligence (AI) and an entirely new approach to fighting cancer saved Carter’s life. A man expected to die three years ago has instead been actively weighing into the US’ recent 2018 mid-term elections, much to the delight of the Democrats and to the chagrin of the Republicans.
Something dramatic is happening in the fight against cancer. A combination of diagnostic tools developed by bioinformatics scientists using AI, along with aggressive immunotherapy breakthroughs, has made cancer less of a death sentence and more of a word puzzle. Now, oncologists no longer speak of ‘remission’; they are beginning to hesitantly use the word ‘cure’.
Dr. Halima Bensmail, a Moroccan AI scientist at the Qatar Computing Research Institute (QCRI), has been instrumental in developing a diagnostic tool for certain types of brain cancer that is also applicable to other cancers—including breast, ovarian, and lung cancer—to identify key driver genes. It can also potentially be used to identify key therapeutic targets in diseases prevalent in Qatar, such as diabetes and obesity.
Together with other QCRI researchers and in collaboration with the University of Sannio in Italy, Columbia University Medical Center, and Henry Ford Health Systems, Bensmail built a machine-learning algorithm that can identify the main regulators of separate brain tumors. Knowledge of the purpose of these driver genes and the status of these cancer subtypes could further support the search for treatments or prognostic information in both glioma and other cancer types. The work reveals the identity and biological activities of the main genes that regulate and characterize the differences between different glioma subtypes.
Bensmail initially studied mathematics at the University of the Sciences in Rabat, and completed her Ph.D. at the University Pierre et Marie-Curie (ParisVI). Prior to joining QCRI, she taught Biostatistics and Bioinformatics at the University of Virginia Medical School and worked as a scientist at the Data Theory Center at the University of Leiden in the Netherlands.
She is convinced that the nascent field of bioinformatics, coupled with unconventional approaches to cancer treatment, will save millions of lives in the future. “With the growth of Big Data research and recent developments in sophisticated high-throughput data platforms, an emerging concept is to prevent or treat diseases that takes into account a patient’s variability in terms of genetic inheritance, lifestyle, and environment,” says Bensmail. “There is an increasing awareness of the value of integrative medical data.”
“Machine learning, a branch within the area of AI, is helping computers to learn from data and discover patterns without being explicitly programmed. For example, we developed a deep learning AI algorithm that could provide doctors with an automatic analysis of the brain structure while the patient is still in an MRI scanner. This would save the physician time, which he or she could use for more relevant tasks, such as patient treatment. Ongoing research, testing, and measurable improvements will be necessary to pave the way for the use of intelligent machines in hospital settings. These machines will not replace doctors but support them in their everyday work.”
“Another example is using classification and prediction algorithms, to check the predisposition of a population—for instance, the Qatari population—to chronic disease such as diabetes or cardiovascular disease,” adds Bensmail. “In this situation, if we collect data in real time (cross-sectional data) and combine it with their historical data (cohort) as well as data from millions of similar patients, an intelligent system could predict a heart attack with a high rate of precision.”
Read Also: Demystifying Cancer in Morocco
Discoveries of ways to harness the immune system to attack cancer have won the Nobel Prize in Physiology and Medicine. James Allison of the University of Texas MD Anderson Cancer Center in Houston and Tasuku Honjo of Kyoto University in Japan each discovered methods of removing the immune system’s ‘brakes’ that prevent it from attacking tumor cells. Following that path, a special branch of cancer immunotherapy aims at introducing T-cells, which have been specially engineered to recognize a tumor. However, in this case it is of utmost importance to ensure that the T-cell does indeed recognize the tumor and nothing else other than the tumor.
If introduced T-cells recognize healthy tissue, the outcome can be fatal. It is therefore extremely important to understand the molecular interaction between the sick cell, such as the peptide (which plays a crucial role in the monitoring of cells in our body by the human immune system) and the MHCI (which binds peptides), and the T-cell.
One way to understand this mechanism is to build a classification model—a machine learning algorithm—and apply it to increase the understanding of the molecular interactions governing the activation of T-cells.
Bensmail and her colleagues at QCRI have developed an algorithm that can be extended to understand the molecular interactions governing the activation of the T-cells using TCGA data (Cancer Genome Atlas database). They are applying it to predict which Master regulator is activated negatively or positively in response to the immunotherapy. They are currently testing this for breast cancer and hope to generalize it to 32 other cancer types that are part of TCGA database.