Tim Dunn

Senior Bioinformatics Scientist @fulcrumgenomics. Views are my own.

[Tim Dunn @T1MD1]

New vcfdist paper on bioRxiv!

Key Takeaways:
1) Jointly evaluating small and structural variants decreases measured FN+FP by 20-50%
2) 43-92% of phasing flip "errors" are FP due to variant representations

More below

Code: https://t.co/v3gz57lEPn
Paper: https://t.co/j9U7sQ4wCu

[Nick Mark MD @nickmmark]

🚨Apparently all NIH Study Sections have been suspended indefinitely.

For those who don’t know, this means there won’t be any review of grants submitted to NIH

Depending on how long this goes on for, this could lead to an interruption in billions in research funding.

[Tim Dunn @T1MD1]

Finally took the time to create a BlueSky account, you can follow me @timd.one

[Tim Dunn @T1MD1]

@i000 @strnr Thank you!

[Tim Dunn @T1MD1]

After nearly six months waiting for review, our extension to vcfdist was finally published in BMC Genome Biology this week!

https://t.co/5tPk1zwJXp

[Stephen Turner 🦋 @stephenturner.us @strnr]

Jointly benchmarking small and structural variant calls with vcfdist https://t.co/C9Qa4CMRVG 🧬🖥️ https://t.co/54xepHNWTa

[Tim Dunn @T1MD1]

@AaronPomerantz @iiSeymour @rrwick @nanopore Cool! With 2/3 errors occurring in the same "GAGCTCC" motif, it makes me wonder if there's some kind of base mod, or if the motif is just underrepresented in the training data.

[Damian Carrington @dpcarrington]

Climate experts were increasingly saying that keeping heating below 1.5C is near impossible, yet it remains the global goal. So I asked hundreds of top IPCC scientists what they thought. What they said shocked even me…
🧵 1/n #ClimateCrisis

[Tim Dunn @T1MD1]

@baobaoyaobaobao @fulcrumgenomics Thanks Emily!

[Tim Dunn @T1MD1]

I'm thrilled to announce that I successfully defended my PhD thesis today, "Improving select applications of long-read DNA sequencing"! After graduating, I'll be joining @fulcrumgenomics as a Senior Bioinformatics Scientist. I can't wait to see what this next chapter brings!

[Tim Dunn @T1MD1]

Excited to see a new benchmarking analysis using vcfdist!

[Zamin Iqbal @ZaminIqbal]

Mandatory reading and surprising results from Michael Hall and colleagues - much better SNP/indel calling results from latest nanopore (beating illumina), and deep learning methods really doing well. Extremely thorough work
https://t.co/Y4oQH1BeOc

[Zamin Iqbal @ZaminIqbal]

Mandatory reading and surprising results from Michael Hall and colleagues - much better SNP/indel calling results from latest nanopore (beating illumina), and deep learning methods really doing well. Extremely thorough work
https://t.co/Y4oQH1BeOc

[Tim Dunn @T1MD1]

@viveksrm24 @fredhutch Congrats Vivek!

[Tim Dunn @T1MD1]

Finally got around to adding vcfdist into bioconda. Installation is now as easy as `conda install vcfdist`.
https://t.co/v3gz57lEPn

[Luca Palmieri @algo_luca]

Google is donating $1M to the @rust_foundation to improve C++/Rust interoperability.

This is big news!

[Matt Holt @holtjma]

Excited to announce that our accompanying paper on HiPhase has been published today, I'll highlight some of the additions in this thread! #PacBio #HiFi

“HiPhase: Jointly phasing small, structural, and tandem repeat variants from HiFi sequencing”: https://t.co/sYFwXLlhcS

[Tim Dunn @T1MD1]

Comparison to prior work: lower measured FP and FN rates than vcfeval or Truvari, and similar results to Truvari refine (WFA/MAFFT). Truvari is faster and works with larger variants, but skips SNPs and the more accurate WFA/MAFFT versions don't scale to large BED regions

[Tim Dunn @T1MD1]

New vcfdist paper on bioRxiv!

Key Takeaways:
1) Jointly evaluating small and structural variants decreases measured FN+FP by 20-50%
2) 43-92% of phasing flip "errors" are FP due to variant representations

More below

Code: https://t.co/v3gz57lEPn
Paper: https://t.co/j9U7sQ4wCu

[Tim Dunn @T1MD1]

As noted above, joint evaluations reduce measured false negative and false positive variant calls across the board: by 28.1% for SNPs, 19.1% for INDELs, and 52.4% for SVs over 50 bases on 3 whole-genome datasets.