FroM Superstring to Indexing: a space-efficient index for unconstrained k-mer sets using the Masked Burrows-Wheeler Transform (MBWT) https://t.co/YagAm5YBz3 #biorxiv_bioinfo

Today we’re excited to freely share an early-version of, perhaps, the world’s most expansive genetics dataset: Logan. #bioinformatics #petabase #genetics #genomics #openscience https://t.co/CQSBmvn7se

A new nanopore sequencer from BGI https://t.co/mnz3LdErL1 Unfortunately, they do not provide reads to evaluate them.

In what journal do you publish? I am looking for well-established bioinformatics journals with reasonable APC, but I can't find anything below $3000 😓

Soon, we will see more long and ultra-long read technologies that combine sequencing, sequence alignment and AI for error correction. I am proud that our lab has contributed to this by https://t.co/hHFbezdGeJ https://t.co/MjKazNVExU and https://t.co/KyR3LPx5rT,

[1/11] Thrilled to announce our latest preprint on function-assigned masked superstrings (shortly, f-masked superstrings or f-MS), which transform masked superstrings into an operationally complete data type for k-mer sets. w/@sladky_on and @VeselyPavel_mff

Very happy to receive the best student's paper award at the Bioinformatics part of BIOSTEC2024 conference 🥳! Work on using ILP to improve metagenomic assembly. Soon on biorXiv. With Rumen Andonov and Tam Truong in the @GenscaleTeam.

Could this be useful in genomics? Two reasons to hope 1) When divergence is low the alignment is near-optimal 2) O(n) in time and O(1) in space => probably very fast if well-implemented 🧵6/6

Switched from random walk to methodic exploration of the matrix: overestimation comes down to O(s)! (but this cannot be formalized as an embedding anymore) 🧵5/6

Is that good in practice -> I tested -> NO. Random walks are very slow to explore the space. But what if we got rid of the randomness ? 🧵4/6

CGK alignment will eventually "find" the optimal alignment after a non-matching character with a random walk. If s is the divergence, the overestimation of the distance is in "O(s^2) with probability 2/3" 🧵3/6

The translation of CGK to pairwise alignment is simple: greedily accept matches between sequences, and if two characters don't match... randomly insert/delete/mismatch! This is what it looks like on NW matrix 🧵2/6

Just discovered CGK embedding by @KoudyMK lab. Very fun paper: aligning two sequences in O(n) time and O(1) space 😯 sacrificing a bounded amount of precision. https://t.co/gMlwdzybMj Summary of the journal club in @GenscaleTeam 🧵1/6

HairSplitter can recover more strains compared to previous methods, while being faster. 🧵 3/3

General purpose long-read assemblers like metaFlye generally collapse close strains. The methods available for now to recover the strains were Strainberry, from @r_vicedomini and @RayanChikhi, and iGDA, but worked only for simple metagenomes 🧵2/3

Want strain-specific metagenomic assemblies from Nanopore reads ? HairSplitter is on bioRxiv https://t.co/C3mA4yYHue 🧵1/3

1) Assemblers do not like structural heterogeneity! Often results in a misassembly. 2) Some misassemblies are completely mysterious and have no clear cause. 3) Assembly can be non-deterministic: re-running an assembly with identical parameters can give a different result. (2/2)

QuickDeconvolution is out! Very happy to have developed that within the @GenscaleTeam. Deconvolve your linked reads data to finally have one cloud per barcode.

🎥 Il reste de la place pour le festival Sciences en Cour[t]s ! 🗓️ 13 octobre 🗺️ Diapason ⌚️ 20h 🎫 https://t.co/EXKWb4M7CS 👋 @DiapasonRennes @UnivRennes1 @UnivRennes_2 @Espace_sciences @FeteScience @FondationR1 @IGDRennes @metropolerennes @GROUPEESRA