Rapid video-based deep learning of cognate versus non-cognate T cell-dendritic cell interactions

Machine learning predictions of T cell antigen specificity from intracellular calcium dynamics

Adoptive T cell therapies rely on the production of T cells with an antigen receptor that directs their specificity toward tumor-specific antigens. Methods for identifying relevant T cell receptor (TCR) sequences, predominantly achieved through the enrichment of antigen-specific T cells, represent a major bottleneck in the production of TCR-engineered cell therapies. Fluctuation of intracellular calcium is a proximal readout of TCR signaling and candidate marker for antigen-specific T cell identification that does not require T cell expansion; however, calcium fluctuations downstream of TCR engagement are highly variable. We propose that machine learning algorithms may allow for T cell classification from complex datasets such as polyclonal T cell signaling events. Using deep learning tools, we demonstrate accurate prediction of TCR-transgenic CD8+ T cell activation based on calcium fluctuations and test the algorithm against T cells bearing a distinct TCR as well as polyclonal T cells. This provides the foundation for an antigen-specific TCR sequence identification pipeline for adoptive T cell therapies.

Fully automated platelet differential interference contrast image analysis via deep learning

Platelets mediate arterial thrombosis, a leading cause of myocardial infarction and stroke. During injury, platelets adhere and spread over exposed subendothelial matrix substrates of the damaged blood vessel wall. The mechanisms which govern platelet activation and their interaction with a range of substrates are therefore regularly investigated using platelet spreading assays. These assays often use differential interference contrast (DIC) microscopy to assess platelet morphology and analysis performed using manual annotation. Here, a convolutional neural network (CNN) allowed fully automated analysis of platelet spreading assays captured by DIC microscopy. The CNN was trained using 120 generalised training images. Increasing the number of training images increases the mean average precision of the CNN. The CNN performance was compared to six manual annotators. Significant variation was observed between annotators, highlighting bias when manual analysis is performed. The CNN effectively analysed platelet morphology when platelets spread over a range of substrates (CRP-XL, vWF and fibrinogen), in the presence and absence of inhibitors (dasatinib, ibrutinib and PRT-060318) and agonist (thrombin), with results consistent in quantifying spread platelet area which is comparable to published literature. The application of a CNN enables, for the first time, automated analysis of platelet spreading assays captured by DIC microscopy.