The goal for this CPRIT High Impact/High Risk grant project is to develop a platform that can be integrated in research labs to study fundamental cancer biology related to CTCs, as well as in clinical settings that profile CTC subtypes to assist cancer diagnosis and treatment.
Texas Tech University’s Dr. Wei Li and collaborators have been awarded a two-year $250,000 High Impact High Reward research grant from the Cancer Prevention and Research Institute of Texas (CPRIT) to develop new technology that seeks to address significant challenges in identifying cancer cells in blood.
Some types of tumors can have cells, called circulating tumor cells (CTCs), that escape into the blood systems of the body and migrate to another organ and start growing there. This process is called metastases, which is responsible for approximately 90% of all cancer-related deaths.
CTCs are rare: only a very small number of these cells exist among billions of blood cells making it extremely difficult to isolate and identify them. In addition, there are several subtypes of CTCs—some more aggressive to cause metastases than others. Correctly identifying those subtypes is crucial to determine the best treatment approach.
However, since CTCs are so rare, it is also extremely difficult to create a database from patient samples (whether from one or from many patients) for conventional machine-learning approaches.
With this grant, Li and his co-investigator, Dr. Jay Lu, will develop a method using a specially-structured microchip to effectively isolate CTCs from blood cells, and then develop an efficient AI model to identify those aggressive CTC subtypes.
Li is excited about the impact this project can have on patient outcomes. “Cancer detection requires the development of engineering tools,” he explains. “With the advances we are making in microchip technology and with AI models, researchers now have the potential to overcome the significant challenges of limited data sources.”
Because large data sets are required to train an AI model, the research team will use cancer cell lines growing in the laboratory to generate a large enough data set to train the AI model to recognize the different types of cancer cells.
After the AI model has been adequately trained using the lab cancer cells, the research team will then address one of the major challenges in identifying CTC subtypes- the difference between cancer cell lines and CTCs from patients.
CTCs isolated from patients are different from cancer cells growing in the laboratory because the environmental conditions in those two are not the same.
Once the AI model has achieved the desired reliability in recognizing lab cancer cells, the team will migrate the AI model from cell lines to CTCs by applying transfer learning to adjust the model with images of patient CTCs captured from the microchip.
The transfer learning method is the bridge to connect lab cell lines with real patient CTCs and overcome the heterogeneity.
Dr. Lu explains, “The project will first train a high-performance AI model with the large amount of cancer images acquired from lab culturing of cancer cells. This pre-trained AI model will then be adapted to patients using patient CTCs. Meta learning strategy will be adopted such that although patient CTCs are extremely scarce, the pre-trained model can be well adapted and give good performance in identifying subtypes of patient CTCs.”
Dr. Li and Lu’s multidisciplinary research team also includes Dr. Yifan Wang from the Department of Mathematics & Statistics at TTU and Dr. Robert Bright from the Department of Immunology and Molecular Microbiology at TTUHSC.