Michael Milton

I've developed a #python package for querying UniProt's new REST API! Maybe the first to fully support the new format. Check it out at #unipressed" target="_blank" rel="nofollow noopener">github.com/multimeric/Uni…. In particular I've tried hard to integrate with Python tooling, giving you great code completion: #bioinformatics

We're advertising for a Bioinformatics Research Assistant to work with me at @WEHI_research to process and analyse data from a range of single-cell and spatial 'omics technologies. Full details of the role & how to apply: wehi.wd3.myworkdayjobs.com/WEHI/job/Parkv…

@ClausWilke @iferres91 @jessenleon @yutannihilat_en @quaternioso @minigamedev @jessenleon You are welcome to join the extendr discord! My understanding is that the CRAN machines now have Rust + Cargo installed so they can compile Rust packages: stat.ethz.ch/pipermail/r-pa….

If I had more time…. Anyone know what the status in on calling #rust code from #rstats? Some #bioinformatics algorithms would be nice to see implemented

🧬💻🔥 We are excited to launch #CuratedAtlasQueryR! millions of cells at your fingertips. A thread 🧵. The harmonized and reannotated #CELLxGENE #singlecell human #atlas can be explored and retrieved as #SingleCellExperiment. Please RT! -> stemangiola.github.io/CuratedAtlasQu…

Pipeline release! nf-core/proteinfold v1.0.0 (Protein 3D structure prediction pipeline) See the changelog: github.com/nf-core/protei…

@braincode Does it work for Linux ⇔ Linux situations? The readme makes it sound like it only runs on Windows.

After all those years #rsync has a worthy competitor?: github.com/google/cdc-fil…

@tgamblin @FluxFramework This is a good summary, I didn't realise Flux was a job scheduler? Each of the many times I've gone to flux-framework.readthedocs.io and read "Flux is a flexible framework for resource management", my eyes glaze over and I go do something else. I wish there was a "Flux vs SLURM" page.

@PMinervini We're replacing SLURM with @FluxFramework on the ~40 clusters at @Livermore_Comp. - Heterogeneous GPU/CPU/storage scheduling - Workflow support - Can also run under SLURM/PBS/LSF - WIP: integration with Kubernetes: youtube.com/watch?v=9VwAcS… More: ipo.llnl.gov/sites/default/…

Folks, assume you had to build a fairly large GPU cluster (100+ A100s) for academic use. Which workload manager would you use and why?

@charliermarsh I suspect that PEP 582 will mostly solve this by removing the need for virtual environments and therefore needing a wrapper script (like poetry run) at all. PDM is the most popular package manager to implement PEP 582, so maybe chuck that in to the benchmark?

I love these tools (Hatch, Poetry, etc.), but I wonder what it would take to make these operations zero-cost?

New Research Officer role available in the Bahlo lab for 2023. Connecting up imaging analysis with genetics. Looking for an image analysis specialist with interest in learning about genetics in a highly collaborative lab. wehi.wd3.myworkdayjobs.com/en-US/WEHI/det…

@jermdemo It has implications for inheritance law

this is important because a lot of full-stack Java devs become President of the United States

@jdidion @geoffjentry True about containers, but this goes beyond that, down to the level of the nextflow integration. Why write two sets of nextflow-on-AWS infrastructure?

@multimeric @geoffjentry Lol…look at how many different ways there are to run a container or host a file system. “Multiple ways to skin the same cat” is the core AWS ethos.

I wonder if the new #aws Omics service is using the open source AGC (aws.amazon.com/genomics-cli/) behind the scenes? It would make sense to develop and road test the workflow system separately in this way, before providing it later as a paid service. #bioinformatics

@geoffjentry @fiamh It even supports snakemake! (aws.github.io/amazon-genomic…). It's really quite a nice system, it would be a shame if it gets replaced entirely by Omics.

@fiamh @multimeric Ah - I was under the mistaken belief the cli only had wdl and nf, so that was a point of (flawed) evidence. Alas!

@fiamh @geoffjentry Another good catch. Although possibly they dropped CWL/Toil because it was their smallest user demographic and don't want to support it commercially. Which wouldn't necessarily rule out AGC internals, although it probably does.

@multimeric @geoffjentry Also doesn’t support CWL via Toil; pretty sure this is a new/different implementation

@geoffjentry Interesting! If that's true, I wonder why they developed two separate workflow integration layers. Admittedly there are some visible differences like the lack of `agc-project.yaml` type config file in Omics, but that may just have been abstracted away.

@multimeric On the WDL slack one of the AWS folks said no when I suggested the same. Although it's possible they misunderstood what I meant.

Well, I'm a total convert to the #podman container engine. Full (?) implementation of the #docker API + CLI, fully open source, super easy to install, built-in support for MacOS via virtualisation, and even installs + runs without root. I think it even beats out #apptainer.

"A year's worth of weekends" @realAaronLun's efforts to re-write key scRNA-seq @Bioconductor/#rstats packages into C++ for use with WASM for the magic that powers jkanche.com/kana/ #biocasia2022

@jermdemo I think the question was just an earnest version of the standard question "how is <TOOL Y which is new to me> better than <TOOL X that I have been using for years>". It was only funny to people who know how hotly debated this particular horse race is.

@multimeric why would you ask a racehorse to judge another racehorse?

The eternal battle between #nextflow and #snakemake continues at #ABACBS22 as the invited speaker from nextflow's parent company is challenged by the audience to compare the two workflow managers. Immediately following is a contrasting presentation which demonstrates the (1/2)

use of snakemake for viral assembly! #bioinformatics (2/2)

A prize to the first person who can successfully allocate a >1 quettabyte (10^30 bytes) vector in #rstats and manage to call object.size() on it! (I take no responsibility for bankruptcy from cloud charges and/or murder by sysadmin that may result from attempting this 😅)