Name: RxRx19
Description: Two kinds of cells are taken into consideration: Human Renal Cortical Epithelial Cells(HRCE) cells and Vero cells. Vero cells are derived from the kidney of an African green monkey. The images repository contains four sets of cells: HRCE-1, HRCE-2, Vero-1, Vero-2; each containing cell images for 26 Plates. Each cell is stained with 5 different stains (channels). Each image is 1024 x 1024 x 1. The metadata can be found in 'metadata.pkl'. The deep learning embeddings can be found in 'embeddings.pkl'. Each row in the pkl has a 'site_id' as described in the metadata schema. The remaining 1024 columns is the embedding for that respective site.
The directory 'notebooks' contain the code and exercises for each day of week 2. The details are as follows:
Day 1: Introduction to the dataset, hands-on experience with manipulating the metadata and deep learning embeddings of the cell images in RxRx19, feature extraction from images using basic Computer Vision techniques.
Day 2: Dimensionality reduction techniques are used in the embeddings data. The goal is find a representation of the data in a lower dimension. The factor models explored were Principal Component Analysis, Matrix Factorization and Autoencoders.
Day 3: The feature set derived from Day 2 is used to train classification models, namely Logistic Regression, KNN Classifier, and Random Forest Classifier. This is followed by an introduction to a simple Convolutional Neural Network, which is trained using the feature embeddings and labels as well to predict if the cell image belongs to 'active' or 'inactive' class. Finally, a detailed comparison between all the four methods are made.
Day 4: Doubt sessions regarding concepts and exercises worked on during the previous days, and preparation for the presentation on Day 5.
Heiser, Katie, et al. "Identification of potential treatments for COVID-19 through artificial intelligence-enabled phenomic analysis of human cells infected with SARS-CoV-2." bioRxiv (2020) (link).
RxRx19: The First Morphological Imaging Dataset on SARS-CoV-2 Virus (link)
RxRx19a COVID-19 Image Embeddings (link)
Techniques to extract features from images (link)
Edge detection in python (link)