Category: Data Analysis and Informatics
Existing computational pipelines for quantitative analysis of high-content microscopy data rely on traditional machine learning approaches that fail to accurately classify more than a single dataset without substantial tuning and training, requiring extensive analysis. Here, we present a deep learning tool that overcomes the pitfalls associated with conventional machine learning classifiers. By combining convolutional neural networks with multiple instance learning, we developed a platform capable of classifying and segmenting microscopy images from a variety of screens with minimal labeling input. The tool works by selecting which experimental conditions are controls, updating the network on the new screening data, and scoring the rest of the screen with the updated network. We show that this approach outperforms several previous methods on both mammalian and yeast microscopy images without requiring any segmentation or feature extraction steps. This method takes a significant step toward automating the analysis of high-content screening data.
Oren Kraus– PhD Candidate, University of Toronto, Toronto, Ontario, Canada
University of Toronto
Toronto, Ontario, Canada
I'm a PhD candidate in Brendan Frey's PSI group. My research focuses on applying cutting edge machine learning techniques (specifically deep learning) to high throughput microscopy screens of cell biology. In collaboration with Charlie Boone and Brenda Andrews at the Donnelly Centre for Cellular and Biomolecular Research (CCBR), I am generating datasets and training models with millions of individual cell objects from genome wide screens in Yeast. Currently I am developing models to predict protein localization and cell cycle stage, as well as unsupervised models to identify morphological outliers and discover novel subcellular phenotypes. I received my BASc and MASc in mechanical engineering at the University of Toronto.