"Franklin was an equal member in a group of four scientists [...] recognized by her colleagues as such, although that acknowledgement was both belated and understated [...] deserves to be remembered not as the victim of the double helix, but as an equal contributor to the solution of the structure."

💡 Bonus: Open Snakemake pipelines for both simulation and imputation are provided:
- github.com/vibaotram/si...
- github.com/vibaotram/im...

Feel free to test and optimize imputation strategies for your own datasets before committing to large-scale LCS projects.

💥 Downstream impact:
- PopGen analyses (relatedness, population structure) were congruent with imputation accuracy and depend on genetic relatedness.
- Inbreeding coefficient estimates were more sensitive to imputation.
- Demographic analysis on imputed data showed similar results across the tools.

🔑 Choosing your imputation tool:
- All tools worked well in highly related populations.
- Accuracy varied in populations with low relatedness, but GLIMPSE2 & QUILT2 generally outperformed the other tools.
- STITCH is useful when no reference panel exists but results in some missing data.

👩🏻‍💻 We tested 5️⃣ imputation methods (GLIMPSE2, GeneImp, QUILT2, STITCH, Beagle5.4), simulated realistic populations with different genetic structures, and applied the methods to data from 283 wild hihi (stitchbird) 🧬.

We explored that in this new study in Molecular Ecology Resources 🔎

onlinelibrary.wiley.com/doi/10.1111/...

@vbtram.bsky.social
@huizhentan.bsky.social
@katarinastuart.bsky.social
@annasanture.bsky.social