High-Content Imaging Data Analysis

Customer:
US-based Fortune 500 pharmaceutical company
Technology:
OpenAPI, Viewers, Apps, Scripting/Python, MVA, Jupyter, statistical hypothesis testing, exploratory data analysis

The goal of this project is to provide web-based interactive analytical capabilities for high-content imaging data processed with the customer’s system developed in-house. The system contains high-resolution imaging data of plates and wells, along with thousands of features on the well level calculated using computer vision algorithms.

Our application connects to the customer’s web service using Datagrok’s OpenAPI integration, and lets the user specify the data of interest and retrieve it along with the selected features across protocols and experiments. After retrieving the data, the application lets users interactively explore results, including performing multivariate analysis, retrieving well images, performing statistical analyses and hypothesis testing, etc. The platform allows the seamless integration of out-of-the-box high-performance viewers with the custom analyses and functions developed in R or Python.

A custom analysis can be saved and shared with others by simply copy-pasting the URL on top.

In the following picture we see application-specific filters on the left, filtered plates in the tiles view, grid with wells data (including features) for the selected plates on top, barchart and histogram next to it, well image for the selected well on the right, and plate visualization on the bottom.

The context panel on the right shows information about the currently selected object. In a picture below, a currently selected object is a well that contains a chemical structure with the specified identifier. In addition to the image of the well being shown, our data augmentation system automatically finds other useful information that could be derived from the current context. For that chemical compound, it shows two additional information panels with some information about that compound from in-house compound registration and plate management systems. Note that the information panels are not part of this high-content analysis solution; they ship as part of the “Data queries” package, and are activated whenever a user clicks on the chemical structure. This is an efficient way of knowledge dissemination, where contextual information gets pushed to the user.

request a demo