Sign in

Roland Faure

@rfaure.bsky.social
78 followers 137 following 29 posts

Sequence bioinfomatician, algorithms, methods. Postdoc in Institut Pasteur in Rayan Chikhi's lab

PostsRepliesMedia
Roland Faure @rfaure.bsky.social · 09/10/2026
Third application: imagine you have sequenced your metagenome but not deeply enough to have a good assembly. Then you can go through the Logan accessions and recruit highly similar sequencing datasets to help co-assemble your sample.
100 1
Roland Faure @rfaure.bsky.social · 09/10/2026
Metadata in the SRA is not always precise and reliable. We investigated all 'organism' tags by checking coherence with tags from most similar accessions. We found thousands of misannotated datasets. We found even more datasets for which we specified the tag ("metagenome" -> "mouse gut metagenome")
100 1
Roland Faure @rfaure.bsky.social · 09/10/2026
First question we can answer with the FracMinHashes: are we exhausting the number of new DNA we sequence? Answer: no in general (as of 2023). In 2023, we were still discovering more and more new stuff every year. However, for gut microbiota, yes: the amount of novelty per sequenced bp is dropping.
110 1
Roland Faure @rfaure.bsky.social · 04/06/2026
We use a new statistical test, which I find beautiful in its simplicity. The idea is to exploit the length of the reads to look at multiple loci at the same time. The p-value of a *group of* variants being due only to sequencing errors is 0.1^(ab)*C(m,b)*C(n,a), C denoting binomial coefficients.
130 3
Roland Faure @rfaure.bsky.social · 04/06/2026
The key is distinguishing SNPs from sequencing errors. In metagenomic data, you cannot rely on "50% of the reads should carry the variant", as you could with human/diploid reads.
100 1
Roland Faure @rfaure.bsky.social · 04/06/2026
SNooPy calls much more SNPs than other tools, including DeepVariant. This is especially true for datasets with high coverage and several strains of the same species, typically human gut metagenomes sequencing 🦠
100 1
Roland Faure @rfaure.bsky.social · 04/12/2025
Sorry for the first figure, got a problem of background, here it is:
000 0
Roland Faure @rfaure.bsky.social · 04/12/2025
Tests show that: 1/ SNooPy has the best recall in our tests 2/ Using genomic long-read SNP callers does not (always) work well: most tools have very low recall, but DeepVariant perform much better than other tested methods 3/ The recall of all tools is still far from 100%
100 1
Roland Faure @rfaure.bsky.social · 04/12/2025
Preprint out! Check out our new long-read metagenomic SNP-caller, SNooPy 😀. Work with Chris Quince. Thread 🧵 👉 www.biorxiv.org/content/10.6...
1138 13
Roland Faure @rfaure.bsky.social · 03/10/2025
🧵5/6 The sketching makes assembly extremely fast: a gut metagenome sample of 138Gbp of sequencing data was assembled in less that 2h and 10G RAM on 8 threads ⚡. And thanks to MSRs, *highly similar strains are not collapsed*
111 1
Roland Faure @rfaure.bsky.social · 03/10/2025
🧵3/ 6 MSRs have been defined by @lblassel.bsky.social @rayanchikhi.bsky.social and @pashadag.bsky.social in pmc.ncbi.nlm.nih.gov/articles/PMC.... Take a sequence, a value of k, and stream all k-mers through a function that output either a base or the empty character, and you got your sketch
110 1
Roland Faure @rfaure.bsky.social · 07/03/2025
So glad to have participated in #DSB2025, what a great workshop! For some mysterious reason it was the first time I attended after 3 years of sequence research. Thanks to all participants & organizers 😃
120 2
Roland Faure @rfaure.bsky.social · 04/12/2024
So glad to have successfully defended my Ph.D. last week 😀 Work on producing haplotype-resolved metagenomic assemblies using noisy long reads (HairSplitter) and high-fidelity long reads (Alice assembler, unpublished yet). Thanks to my advisors Dominique Lavenier and Jean-François Flot ❤️
193 9