Menu Close

ISMB 2022: scPolyphony

I had the great privilege of attending the Intelligent Systems for Molecular Biology conference in Madison, July 10–14. Here, I’m sharing about something interesting I learned there!

If you’ve spent time in the bioinformatics or genomics space recently, you’re probably familiar with the advantages and challenges of single-cell sequencing! To race through some of them, single-cell sequencing data can be used to uncover novel cell types, refine cell type biomarkers, and disentangle cell composition from gene expression quantification — but the computational aspect of determining cell type can be demanding, and significant variation between experiments can be present due to technical effects as well as real biological differences.

scPolyphony is a new tool (preprint here!) that helps improve clustering accuracy by allowing human and AI input into matching a new dataset onto a reference. Anchors — computationally-matched groups of cells — are defined for each dataset, and the researcher is given the opportunity to add, delete, or modify those anchors if a mismatch appears to have been made. This interactivity is the most unique aspect of this tool, in my opinion, and really increases its usefulness.

Image from the Polyphony publication demonstrating embedded clustering before and after manual refinement of anchor-points and cell-type predictions

Unlike the majority of open-source bioinformatics tools, this can be set up as a web server by someone with IT experience and then used through a web portal by a scientist with less computing expertise, which as a core manager makes me excited about the possibility of integrating researchers’ biological knowledge into the analytic process more smoothly. I’m planning to test it on some public data and will hopefully be able to post some examples and a tutorial in the future!

Leave a Reply

Your email address will not be published. Required fields are marked *