Rishi De-Kayne @rishidekayne.bsky.social · 06/04/2026Our phylogenetic analysis shows this clearly: split domains from tandem proteins land in the same clusters as their separately annotated counterparts, not together as you'd expect if they shared a single evolutionary origin. [7/9] 100
Rishi De-Kayne @rishidekayne.bsky.social · 06/04/2026This matters because protein structure prediction is increasingly being run on sequences from non-model organisms where we have little prior knowledge. Erroneous sequences from genome misassembly can silently percolate through prediction models into structure databases. [8/9] 130
Rishi De-Kayne @rishidekayne.bsky.social · 07/04/2026All code and data are available on GitHub: github.com/Arcadia-Scie.... We hope this work helps others think about quality control of input sequences before running structure prediction at scale. @arcadiascience.com [9/9]github.com 000
