@HighFreqAsuka @marikgoldstein You also don't really need to know stochastic calculus for what most people consider "diffusion models" these days (i.e. where the diffusion process is learned via a neural network).
English
Julian Hess
866 posts
Julian Hess
@JulianMHess
Computational Biologist//@broadinstitute
Katılım Haziran 2019
131 Takip Edilen165 Takipçiler
@marikgoldstein You don’t really need to know stochastic calculus for most finance jobs.
English
diffusion models are just a ploy by CS PhDs to get their departments to pay for them to learn stochastic calculus so that they can get finance jobs
English
@NmrReyez @GenomicsCow Helical chirality may not matter per se, but knowing nucleic acids chemistry will be important always.
English
@GenomicsCow > - which way the helix turns
this will be important never
English
No matter how high up you go or how far you branch away from academia, try to remember
- stats
- variant calling
- how PCR works and its various biases
- fluorophores and imaging
- which way the helix turns
…What else is universal for genomics?
claire vo 🖤@clairevo
Hard skills matter.
English
@GenomicsCow Knowing the full stack. Be able to explain at a graduate level sample/library prep, sequencing, read processing, etc., all the way to cohort-level analyses. But most importantly, know the underlying cell biology motivating those analyses.
English
@NateSilver538 Could there be less variance between polls due to the addition of recall weighting?
English
Spotted the #Auroraborealis on my flight a couple days ago. The low light capabilities of modern phone cameras are astonishing; this is an image that literally could not have existed even 10 years ago.
English
going hard in the paint this weekend. feels good, like real work should.
English
There are few feelings more satisfying than merging a huge branch into master and watching this happen.
GIF
English
@GenomicsCow Not quite as bad a strategy as competing with customers; this is only bad for ILMN if those partners explicitly only supported them.
x.com/JulianMHess/st…
Julian Hess@JulianMHess
@OmicsOmicsBlog Risk of too much vert. integration is competing with partners. E.g. til now 10x had only officially supported ILMN, but the Fluent acquisition will likely change things. ILMN needs the consumables revenue driven by 10x customers more than 10x needs ILMN as a partner platform.
English
The old message “We compete with our clinical customers”
The new message “We compete with our library prep partners”
Keith Robison@OmicsOmicsBlog
Illumina Would Like to Change the Conversation omicsomics.blogspot.com/2024/08/illumi…
English
@OmicsOmicsBlog Risk of too much vert. integration is competing with partners. E.g. til now 10x had only officially supported ILMN, but the Fluent acquisition will likely change things. ILMN needs the consumables revenue driven by 10x customers more than 10x needs ILMN as a partner platform.
English
Illumina Would Like to Change the Conversation
omicsomics.blogspot.com/2024/08/illumi…
English
@GenomicsCow If they’re going to go down the road of spatial barcodes, would it be cheaper to license SeqScope?
But given the size of their flow cell, I wouldn’t be surprised if they’re also looking into in situ tech à la Singular.
English
@Carnage4Life Pareto in action. At most companies, ≥80% of the work is done by ≤20% of the people. Even more so in tech — probably closer to 95/5. Big tech is rich enough that they find that 5% by indiscriminately hiring them along with the other 95% and touting this as “growth.”
English
@GenomicsCow The real question is, if 23andme offered WGS, would it provide a sufficiently richer report to rekindle interest in the product? A 🐓/🥚: maybe the report would only be richer if enough people contributed WGS genotypes…
English
@GenomicsCow Only so many traits a SNP array can elucidate, most of them mono/oligogenic and silly (“so that’s why I hate cilantro!”), and enough of the population has taken one that all of the sordid family secrets have been revealed.
English
Who is taking a DTC test like 23andme in 2024? It seems like we have collectively stopped talking about it at the dinner table
English
@BiologyMusings Exactly this. I think the “on-flow cell library prep” they’ve been teasing will leverage Fluent’s templated emulsion IP.
English
It feels like a patent move for sample preparation rather than single-cell analysis. Keep in mind, for every $ 10x makes, Illumina earns $1. They've made over $1.5 billion to date without lifting a finger.
English
Genentech published a sequencing comparison of all commercial single-cell technologies two weeks ago. Fluent offers a low cost per cell but has a long way to match 10x in sensitivity.
ILMN could go for 10x/Parse, probably didn’t want to upset their friends at the FTC..again
English
@geoffjentry Agreed that it's hard to gauge this with specific interview questions. I try and assess more generally whether candidates are curious and detail-oriented enough to be aware of/give a damn about what many would dismiss as esoteric minutiae.
English
@JulianMHess agree & disagree. This specific example, people *should* know. But it's easy to forget in the moment & especially in interview situation. Also, linters pick up on it right away, so easy to have it be something offload to the robots.
English
Hot take: knowing trivia like this often saves hours of downstream debugging and tech debt, and is a distinguisher of programming ability. A good programmer would be able to quickly debug issues caused by footguns like this; a great programmer would avoid them in the first place.
RiscV@MKVRiscy
There’s a category of “programming language trivia” like this interviewers love to ask about that I think people put too much importance on If you’re writing code that requires such a deep understanding of the Python interpreter that you can explain this, you’re writing bad code
English
… and a poor programmer would let this exact issue contribute to the collapse of a major social news platform.
x.com/robcobbable/st…
Rob Cobb@robcobbable
@MKVRiscy reminds me of @Lethain’s war story from the digg v4 launch lethain.com/digg-v4/
English
@jermdemo @OmicsOmicsBlog To be fair, they placed two instruments; it just appears that one of them wasn’t purchased.
English
@OmicsOmicsBlog i rather appreciate the "an instrument" phrasing - they are not trying to obfuscate that as if they sold multiple sequencers
English
Singular Genomics Records $700K in Q2 Revenues | GenomeWeb genomeweb.com/sequencing/sin…
Català
@OmicsOmicsBlog @nilshomer @sinabooeshaghi @biomodalhq But as these workflows become more common, our venerable legacy pipelines must evolve to gracefully accommodate them.
English
@OmicsOmicsBlog @nilshomer @sinabooeshaghi My boomer opinion only holds for legacy workflows with few-to-no pre-alignment steps worthy of PG header lines. Not the case for cutting edge workflows with more pre-alignment analysis (e.g. spatial coord. extraction, UMI/duplex cons. calling, @biomodalhq meth. calling, etc.)
English
