Bryce Kille
18 posts
Bryce Kille
@BKille
PhD student and former software engineer in the Treangen lab at Rice University. MS in computer science from UIUC.
Houston, TX Katılım Aralık 2019
414 Takip Edilen144 Takipçiler
This is foundation work of theory and practice of minimizers — a class of randomized sampling algorithms to sample kmer from strings. I encourage everybody working on bioinformatics to read it! Congrats 👏🏻 to the authors!
Bryce Kille@BKille
In a new preprint, @curious_coding and I led the derivation of a new lower bound, g’ (red), on k-mer sampling scheme density: - our bound appears to be tight for k ≡ 1 (mod w). - existing schemes are nearly optimal! - mod-minimizers are optimal for large sigma for k ≡ 1 (mod w)
English
@curious_coding @giulio_pibiri @traingene @SapovalNicolae Lastly, having @curious_coding as co-lead has been a blessing. In addition to being a TeX wizard, his attention to detail and ability to communicate concepts so clearly is impeccable. A great collaborator all-around 🤝
English
@curious_coding @giulio_pibiri Our thanks go out to the superstar Rice undergrad coauthors, Drake McAdams and Alan Liu, for spending their summer working on this, @traingene for supporting this theoretical detour, and @giulio_pibiri and @SapovalNicolae for their feedback (:
English
In a new preprint, @curious_coding and I led the derivation of a new lower bound, g’ (red), on k-mer sampling scheme density:
- our bound appears to be tight for k ≡ 1 (mod w).
- existing schemes are nearly optimal!
- mod-minimizers are optimal for large sigma for k ≡ 1 (mod w)
bioRxiv Bioinfo@biorxiv_bioinfo
A near-tight lower bound on the density of forward sampling schemes biorxiv.org/cgi/content/sh… #biorxiv_bioinfo
English
@curious_coding We show that existing sampling schemes are actually a lot closer to being optimal than previously known. Thus, we encourage the usage of the best-known schemes so far. Making these more efficient is ongoing work.
English
@curious_coding We use our new lower bound to guide an ILP in identifying optimal schemes, and indeed, our lower bound appears to be tight when k ≡ 1 (mod w)! For other w, k, especially when large, our bound is nearly tight.
English
Here's a write-up on minimizers and my new idea!
curiouscoding.nl/posts/minimize…
For k>1.25w, these 'reduced minimizers' get surprisingly close to (1.5+2/w)/(w+1), supposedly(?) the lower bound.
Ragnar {Groot Koerkamp} 🦋@curious_coding
For anybody using (l=31, k=21) minimizers, I have an exciting announcement coming up as soon as I finish writing stuff down. Saves 10% density using 2 if statements!
English
@curious_coding @krsahlin The t-minimizers would be a subset of (w, t)-minmers, right? Since it is picking the rightmost minmer for each window.
English
@krsahlin Ah thanks for the reference!
Minmers look similar but slightly different. I'll implement them and add them to the benchmark.
English
@nomad421 Thanks, it has come a long way since then, so I hope you enjoy!
English
It’s been an honor to work on such a popular tool and meet so many great scientists in the process. Thank you @aphillippy for the opportunity! I couldn’t have asked for a more ideal project, it was the perfect balance of algorithmics, probability, and high performance computing.
Adam Phillippy@aphillippy
Our latest methods paper "Minmers are a generalization of minimizers that enable unbiased local Jaccard estimation" (aka MashMap3) is out with @BKille @erikgarrison @traingene! There's an interesting backstory here 🧵 \ biorxiv.org/content/10.110…
English
With all of this wrapped up, I’m looking forward to whatever comes next, especially if it involves hopping on the Rust bandwagon (:
English
While I expected to learn a lot about MashMap-related content over the last year, I truly underestimated the sheer amount of general research skills I’d take away from the project, particularly thanks to @aphillippy's mentorship, as well as @traingene's and @erikgarrison's.
English
@lh3lh3 @aphillippy @erikgarrison @traingene That's not really a strong argument though, since afaik there are no proven ties between 1-locality and Jaccard bias. It would be interesting to see a general relationship between locality and Jaccard estimation bias.
English
@lh3lh3 @aphillippy @erikgarrison @traingene Thanks! We haven't looked at syncmers, however since FracMinHash was shown to be biased, I'd guess that syncmers are biased too since they are also 1-local.
biorxiv.org/content/10.110…
English
Our latest methods paper "Minmers are a generalization of minimizers that enable unbiased local Jaccard estimation" (aka MashMap3) is out with @BKille @erikgarrison @traingene! There's an interesting backstory here 🧵 \ biorxiv.org/content/10.110…
English
Bryce Kille retweetledi
Happy to announce an open postdoctoral research scientist position in my group @RiceCompSci focused on exciting research questions related to metagenomics & protein structure prediction. Joint position with the amazing @KavrakiLab
Interested? apply today! careers.iscb.org/jobs/view/8304
English
Bryce Kille retweetledi
The problem was really fun to work on!
And as someone knowing little about genomes at the beginning, I was blessed to have invaluable support & guidance from @traingene @beleafsTX @BKille @SapovalNicolae @fuyilei96!
And thank you very much @RiceCompSci for the spotlight! 😁🙏
Rice Computer Science@RiceCompSci
Rice CS PhD student Tiancheng Xu has developed a new hardware accelerator that speeds up genome variant analysis, processing #SARSCoV2 data in minutes. He recently presented this new tool at #FCCM22 alongside his advisors Alan Cox and Scott Rixner. bit.ly/3zNdypw
English
This was by far the largest and most collaborative project I’ve ever worked on and what a rewarding experience it was. Thanks to everyone who helped make it happen!
Todd J Treangen@traingene
Very happy to share the latest paper from my group, SeqScreen, just published in @GenomeBiology. SeqScreen was led by my @RiceCompSci PhD students @AdvaitBalaji & @BKille, spanning 5 fruitful years of collaboration with Signature Science LLC @SigSci genomebiology.biomedcentral.com/articles/10.11… 1/n
English
English
I am very excited to share that Bryce Kille (@BKille) has joined my group at @RiceCompSci, Software Engineer extraordinaire with B.S. from @IllinoisCS, developer of RODEO2 (github.com/the-mitchell-l…) & contributor of C++ bit-algorithms (github.com/vreverdy/bit-a…), and now on twitter!
English