Using Zegami’s unique visualization and unsupervised machine learning to gain new insights high content screening experiments

In December 2021 Zegami delivered an integration project for a global pharmaceutical company. The challenge was to integrate their existing Columbus infrastructure (based on OMERO) into a machine learning based image analysis platform to analyse and integrate data from drug studies, high content screening imaging and functional genomics.

The solution provided an easy-to-use platform that allowed users to leverage machine learning to identify interesting plates/wells and overall trends in the screen based on the data.

The goal of the screen was to find small molecule candidates that increase energy expenditure in adipocytes.

Zegami was instrumental in allowing to compare our image analysis results to the unbiased approach of the image similarity metric and by doing so validate the usefulness of the approach for our screen. Furthermore, by looking at the clustering results we were able to separate the potential hits from the toxic compounds which gave similar readouts in our image analysis results.

Zegami worked closely with us to adapt their platform to our high content imaging needs.

Senior research scientist – Imaging specialist

Zegami worked closely with the team to understand the user requirements and implemented an integration into Columbus. The result was a system that could scale to a large number of screens that combined unsupervised and supervised machine learning techniques with imaging and visualization tools. Using these tools allowed scientists to quickly identify QC issues and new discoveries with their data.

Possible hit

Positive control

Negative control

Toxic