Deep learning for analysing immune cell interactions was my research project for my final year at the University of Glasgow. It was developed over a period of ~7 months from September 2019 to April 2020. This project is fully documented in /dissertation/ and a PDF of the dissertation is available here.
This repository includes an implementation of a convolutional autoencoder and a deep regression model that were developed to research whether we could quality/quantify interaction from images of immune cells (T cells and dendritic cells).
- Can we find an underlying structure in the images of immune cells under different experimental conditions?
- Can a deep learning model be trained to 'quantify interaction' from an image of immune cells?
I tried to find cluster of images of T cells and dendritic cells around their level of stimulation: Unstimulated, stimulation by OVA peptide, and stimulation by ConA.
Our 'quantity of interaction' was amount of overlap between the T cells and dendritic cells.
.
├── data <-- find sample data, evaluation data, and notebooks
├── dissertation <-- latex files and figures for generating my dissertation
├── meetings <-- meeting minutes and powerpoint presentations
├── presentation <-- final presentation slides
├── src <-- code of the project
└── status_report <-- status reports submitted after summer and before christmas
Instructions for running the code can be found in src/README.md. If you want to explore the code you can look at src/manual.md.