Taking feedback from customers is the best way to improve any product, a practice we take seriously at Elucidata. Recently, while chatting with one of our customers, we realized that across the drug discovery process, researchers use public data to build and validate hypotheses.

Researchers fetch data from multiple public data sources such as TCGA, UK BioBank, cBioportal, etc., & perform the meta-analysis. More often than not, they analyze experimental data differently than what has been published.

One critical feedback we received from the researchers that use Polly, especially the Bioinformaticians working on different omics datasets, was the repeated need to group samples from multiple OmixAtlases such that it becomes easy to analyze data from different datasets/across repositories.

So voila! Here we are with the incredible ‘Cohorting’ feature for Polly that helps group data.

Put simply, you can now group datasets or samples based on metadata of interest on Polly.

How Exactly Does This Help Researchers?

Cohort studies are a powerful tool for any multi-omics comparative analysis. For example, researchers can look at baseline multi-omics data for people without a particular disease and examine the factors that differed between those who developed a condition and those who did not.

Moreover, you can build groups aka cohorts, by combining datasets that can help you identify patterns/gene expression imbalances and so much more. The possibility of detailed analysis across multiple datasets will allow users to reach actionable insights so much faster!

What Does Cohorting on Polly Allow a User to Do?

A user can use the cohorting feature to do one or more of the following:

• Based on the metadata properties, load samples or datasets in to a cohort;
• Merge dataset, sample, feature level metadata and data matrix for all samples in cohort and create analysis-ready data frames;
• Do QA for samples based on sample level metadata or data matrix info in the cohort bundle;
• Remove & add samples from a cohort bundle;
• Move cohort bundle from one workspace to another;
• Enable cohort creation across all repositories with .GCT files where one dataset has one or multiple samples (currently applicable to the two datatypes: Mutation and Transcriptomics).

Here’s a video to give you a glimpse of how cohorting using these features looks on Polly Python!

Which New Features Are Coming for Cohorting on Polly in the next Quarter?

We’ve also included these exciting features to our To-Do to enrich user experience:

• Auto-generate a report of samples
• Edit metadata information (and .GCT file) in the cohort bundle
• Link the cohort updates with the curation application

Which Features Are in the Future Pipeline for Cohorting on Polly?

Last, but not the least- few more that we would love for our users to have in the future:

• Updating the .GCT file
• Enable cohorting in repositories where 1 dataset is tagged with multiple samples
• Combine individual .GCT file to create a master .GCT file for downstream analysis
• Demonstrate within & between cohort data analysis using a python notebook
• Clone an existing cohort bundle
• Fetch different versions of an existing cohort bundle for analysis against one another ( Retrieve an older cohort bundle version to work with)
• Reduce the time taken to create a cohort bundle
• Export the cohort to an Enterprise Omixatlas
• Across repositories standardization of sample and feature level metadata used in the consumption of a given datatype
• Search within a cohort

Reach out to your CSM, if there’s a particular feature that’s important to you and you’d like us to prioritize.

What Is Not Part of the Feature List for Cohorting on Polly?

• Helping users identify which exact sample IDs to be added in a cohort.
• Helping users perform downstream analysis using the master .GCT file.

Can I See a Demo of This Feature?

Yes, Absolutely! Here’s a link to a demo; If you need more help, just reach out to your Customer Success Manager, who’d be more than happy to set up a call to take you through the feature one on one. Or just click here to book a meeting.