NEC ventures into the drug discovery business: pursuing the possibility of developing vaccines for cancer treatment through AI technology
March 7, 2017
──Is it difficult to find the right peptides?
The body has an immune mechanism for inhibiting cancer cells, and white blood cells called killer lymphocytes play an important role in this mechanism for attacking cancer cells. The substance responsible for activating this killer lymphocyte is a peptide consisting of around nine amino acids.
There are 20 natural amino acids. In other words, there is an enormous number of possible activating peptides–20 to the 9th power (approximately 500 billion). In addition, white blood cells have types called HLA, wherein killer lymphocytes are activated only when the administered peptide vaccine matches the patient's white blood cell type.
Using conventional methods of repeating prediction and experimentation, finding a vaccine that is truly effective to each patient requires an enormous amount of time and money. Finding a peptide that induces immunity for multiple white blood cell types is thus like finding a needle in a haystack. The method for discovering them, therefore, has been a major challenge. Generally, after selecting candidate peptides by predicting the binding to the HLA molecule, each candidate is subjected to binding and immunity activation experiments, repeating the entire series of processes until a good candidate is found. Using this method for discovering a peptide that can activate immunity for multiple white blood cell types was almost impossible.
This is where we applied "computational science" in lieu of "experimental science" as a method for discovering potential peptides. Using Artificial Intelligence (AI) technology, which NEC has been pursuing as a key technology for the future, we attempted to drastically improve peptide predication accuracy. In particular, we developed a unique "immune function prediction technology," wherein machine learning actively provides instructions on the succeeding experiment and then learns its results (called "active learning"). This enabled us to discover potential peptide vaccines against particular antigens at an accuracy of 93%, from approximately 500 billion possible combinations using only 200 experimentations. This "immune function prediction technology" is based on the principle that "increasing the accuracy of AI prediction would make it possible to accurately find peptides that activate immunity across multiple white blood cell types, even for unknown cancer antigens." We have been engaged in joint research with Yamaguchi University and Kochi University since 2013, eventually leading to the establishment of a new company that promotes the development and application of therapeutic cancer peptide vaccines discovered using NEC's advanced AI technologies.
Improving the capability to process massive amounts of data using AI and increasing the prediction accuracy using active learning are groundbreaking milestones that far exceed what traditional methods can achieve. These achievements were made possible through the combination of the medical knowhow we have accumulated through the years and NEC's cutting-edge AI technologies. I believe that we are on the path to creating a new type of drug.
──In particular, what are those achievements, and what is the role that CYTLIMIC aims to fulfill?
We were able to find useful peptides from two antigens in particular, namely HSP70 and GPC3. The peptides were found to be compatible with the genetic profile of approximately 85% of Japan's population. Patents for the two peptides have already been applied for. In addition, we have also discovered adjuvants that enhance the immune function of the two peptides. We have also submitted patent applications for these adjuvants. Moreover, we have already carried out clinical research on these peptides and adjuvants against solid cancers, such as hepatoma, colorectal cancer, esophageal cancer, and pancreatic cancer, and have proven that the peptides activate immunity to a certain extent. CYTLIMIC will move ahead with the development of the peptide vaccines used in the clinical research carried out at Yamaguchi University, and will proceed with conducting various nonclinical and clinical trials and investigating their commercialization in collaboration with pharmaceutical companies.
NEC, which owns 39.9% of CYTLIMIC, will also continue to support the search of candidate substances using AI, not only for cancer peptides, but also for the development of medicines that enhance the immune system against infectious diseases and other human ailments, through academia-industry collaborations.