Life science software pitch
Sales: “this is our product”
PI: “I’ve a phd student who can build something like it”
Sales: “what about when this student graduates?”
PI: 🤷♀️
Sales: “what would that mean for your lab?”
PI: “I’ve a phd student who can build something like it”
“It’s just a few lines of code, right?”
This is the well-known phrase for someone disregarding the effort put into software dev.
Here are other insults for other functions in a genomics company 🧵:
Ex-Illumina CEO Francis de Souza to join the board of directors at Ultima Genomics, a striking strategic play from the up-and-coming sequencing firm that would benefit from his experience and connections
What do you think?
So, the once behemoth
#invitae
has fallen. Over the years I have probably tweeted about them more than any other company, challenging their bold business model.
A mixture of these helpful and snarky tweets in a timeline:
🧵
I’ve heard that the next wave after genomics 🌊is the proteomics 🌊
Here’s why I’m not sure about that:
- What made genomics big? It wasn’t just *one* platform like illumina. It was a series of platforms eg: gels, PCRs, CE seq, NGS.
Which is the most important instrument in illumina’s history? I would argue the HiSeq, which powered them ahead of SOLiD and gave them dominance in genomics centres. The GAii was good but HiSeq I think was what gave them the lead in NGS they never relinquished
The next spatial transcriptomics study should be less on comparison between platforms but comparisons between cell boundary stains and algorithms across different cell types and tissues
Poor segmentation > great sensitivity/specificity
Beware firms boasting their new tech’s publication record
Eg:
- berkeley lights and nanostring claim a large publication list but many are simply citations in a review
- vizgen lists studies that are performed in the original merfish lab, doing stuff you can’t do on a merescope
10 years ago on this month illumina launched HiSeq X Ten, a $10M investment for $1000 genomes. This was illumina at its show-off best:
1. Had to buy ten boxes, not just one, even though the instruments weren’t really dependent on each other or truly connected.
Be weary with spatial transcriptomics: I bet most variation in results at this stage are still related to tissue sectioning and how well it’s RNA was protected prior to it being assayed.
Genomics (NGS) is eating the life sciences world
✅proteomics (readouts for immunoassays/aptamers)
✅histology (spatial)
💰💰And the big grants, of course!
But not just that…
“There must be a bioinformatics tool for that out there already”
There probably is one. But not properly supported nor easy to integrate. One says this when one assumes it’s free to use, right?
Respect bioinformatics
Challenge for spatial biology in 2023: Less focus on pretty pictures. Less experiments just to confirm cell locations. More real experiments Eg treatment v control comparisons, measure gradients of change and with replicates. Prove to be a mainstay technique in life science!
Some good examples of multi omics in this article showing the future
Today however, multi omics to me often reads like
“we had these samples and we ran every technique we could afford on them; the results are presented here sequentially and as single variates” 🐮
This excellent paper has a somewhat subdued writing tone but has good content for all those ppl enjoying the spatial platform comparison studies
Some thorts on results not covered by the authors. . .
Some acquisitions that paid off:
1: illumina / solexa. Easy to forget that illumina was an array co before that. Solexa could have been picked up elsewhere.
I really wonder about standalone software firms
Some areas like LIMS etc seem to do quite well
Some areas like selling analysis seem to really struggle
I feel like many life scientists don’t budget for software so many analyses software have to be bundled in with reagents
Selling panel products:
Sales: “Here is our panel, carefully curated to your research area🥰”
PI: “But can I customize it?”
..months later..
“Ok lots of dev effort went into this but now you can fully customize it🥰”
“This is a bit overwhelming. Can you help me customize it?”
So it might be trendy to 💩on illumina as we don’t like monopolies
Sorry to state the obvious but think of the patients: at this stage bringing genomics into the clinic benefits more from a concerted effort from a large dominant company that benefits from expanding the pie
I’m curious to know whether
#genomics
professionals here on twitter have got their DNA arrayed/sequenced in a non-medical setting.
Why? Why not?
How? Through DTC? (Or Self-performed? 😂)
I wonder if we will look back and wonder how we did ffpe for so long without normalizing results
#spatialRIN
I get it, researchers work with whatever they get access to. It’s just crazy how little we acknowledge how much sample quality changes results
It seems that as soon as a product is available for FFPE tissue, folks try to find the oldest block in their possession to try out. Saw it yesterday for
#Visium
@10xGenomics
and many times with
#TSO500
@illumina
One of the hang-ups I think academics have in moving into commercial roles is losing their gravitas of being a subject matter expert in a field of research
It takes some humility to be able to start asking customers for their opinion vs being the person who gives the opinion
The underlying messages from Roche:
This clinical sequencing market is worth it
We haven’t missed the boat
And we aren’t missing the boat for the next two years either
The point about illumina sequencing is that it derisks a new method
Want to try olink? Well at least you know the read out even though it’s not digital
Want to do spatial? You know what you’ll get with visium without having to evaluate the new imagers out there
I’m Asian and a foodie so I will dare make this simple analogy. Illumina sequencing is like rice. It’s a staple. It’s basic, standard and robust. We know exactly what we are getting. Making it fancy depends on what you have with it but it goes well with almost anything.
Current messaging in spatial:
Nanostring: 6000 plex! And our data doesn’t really suck that bad.
10x: Hey we are 10x! And we have George Church (and his patents)!
Vizgen: We can help you do whatever the Harvard MERFISH lab does.
Akoya: TME TME TME… oh and we will also do RNA too
I don’t have any vintage genomics apparel to wear for
#agbt23
(to look like I’ve made it)
Does it count if I just wear t-shirts from
#agbt22
sponsors who are no longer with us? 🤷♀️🤷♂️🐮
What could disrupt illumina..
- sequencing in situ
- sequencing proteins…
- …or large, customizable protein panels (& 1:1 same between batches)
- long reads w NextSeq read counts/prices
- super cheap sequencer for every lab w ability to impute to references
..in 2025 maybe?
If we’re all here on this platform in a couple of years’ time I fear that I’ll be forced to change my moniker to proteomics-cow. Or at least, multiomics-cow 🐄
What I like about ARK is they have the guts to make calculated predictions for the future we all hope to live in (even if we disagree here and there)
And they have skin in the game.
I suspect in the next few years, there will be a market for risk-stratifying cancer screening cohorts for clinical trials. Hereditary variation, both mono- and polygenic, is one of the most powerful tools to limit over-treatment and improve cost-effectiveness of screening.
Seeing spatial transcriptomics platforms argue over who has better data. Well, at this stage methinks cell segmentation is the great leveller. A bit like NGS when we didn’t know which read aligners to use
1/2 Spatial vendors have the luxury of customers who are willing to buy platforms without seeing all specs in play, let alone demonstrating them first.
This is state-of-the-art stuff. Many of you would have seen them at AGBT
Of course there are use cases
But I’m curious though to see if this really drives even deeper, granular research into the individual cell across the field in general
Molecular Pixelation, developed by Fredriksson and colleagues
@PixelgenTech
, is an optics-free method that uses DNA-tagged antibodies for assessing cell surface protein location at single cell resolution.
So, who from the genomeweb index *hasn’t* been in the news about layoffs since Covid?
Biorad
Qiagen
Fulgent genetics
Oxford nanopore
Bruker
Biotechne
Who else?
In situ Spatial transcriptomics is an interesting field to watch develop. But unlike what I first imagined, it’s not quite like the emergence of NGS.
Some thorts:
If you’re an academic attending
#agbt23
and you paid for your scientific pass you’re doing it wrong.
Next time, as soon as your abstract is accepted ask the company from your methods section to sponsor you.
“Can you make this graph here for me again, but using these colors?”
Your bioinformatics team may know how to use R, but they aren’t your graphic designers.
Respect bioinformatics
This lab is really something
But I wonder how many listen and think, yeah I’ll just run this on one of the other platforms, eg xenium
Also ironic, that you can’t do some of this work on the Vizgen Merscope either
Talk by Xiaowei Zhuang from
@HarvardUniv
on spatially resolved single-cell genomics and making a cell atlas of the human and mouse brain
#singlecellgenomicsday
1/9:
#Germlinediagnostics
: more samples = lower cost and improved services. Lots of labs will benefit from this growing market. But can the market be consolidated by ONE lab? Full length thoughts here:
#Invitae
$ILMN
I wish their alumni good luck. Clinical genomics needs dreamers to challenge cynics like me.
#Invitae
were a Moses company - a company that saw the promised land, but died before it could bring us all there. Someone else will take us there.
#AGBT24
: Who are N26Tec?
“This new method will revolutionize NGS sample prep workflow universally across virtually all assays. The simplicity and innovation will drastically improve workflow efficiency and data quality”
Guess the company doing this
#agbt24
review
“Welcome back! How was AGBT?”
“Awwwesome!”
“How so?”
“Parties galore. Like wow. This was our 1st time so we were blown away”
“Well you asked us for marketing budget. What did you come back with?”
“I got these giveaways from NEB. LOL”
Thanks for sharing!!
My 🐮 thorts on a selection of slides:
1- CosMx FOVs not stitched
Silly development oversight given that there are tools available to integrate with large benefit to results
Users could trim edges but will lose data
For those interested in reviewing our observations, the slides can be found on our website under "Resources." We welcome constructive feedback and suggestions for analysis.
Good post. Worth a second read
Interesting that this solves a problem we whinge a lot about but never got round to solving and kinda accepted the consequences
Worth following to see what use cases come of it
Here’s a prediction: although the hype now is over imaging
#spatialtranscriptomics
techniques with big capex revenues and beautiful imagery, sequencing methods will be the work horse that most individual research groups use.
Samples (especially old ones) should be QC’d to show how their prior-to-you-getting-them conditions impact your assay results, especially for RNA and protein
But we have seen that neither guidelines nor products in this area utilized much
PGx really needs to take off for depression meds. Trial n error of these drugs is long process taking weeks to recognize if it is working and also weeks to wean patient off, before making a change
Presenters, please:
One figure per slide, please
Unless it related to other figures on the same slide, please
Don’t put a bunch of figures in your slides if you don’t talk about them, please
Curious if anyone knows the history. Given library prep was always expected to make up an increasing % of total cost, how did illumina cede their lead to NEB and Kapa and whoever else?
Pro tip from your friendly legal team: a) Avoid passing around rival patents 🤫 b) Refrain from plotting "smackdown plans" on competition 📷 Trust us, it's not a good look in court!
@dr_alphalyrae
There isn’t really a wallet for bioinformatics tools for a lot of labs. I thort there was a good article to reference about this someplace but can’t find it now.
Good enough not perfect
The customers see different flaws in the product than you
If not accountable for revenue don’t be the decider on features
Ok to be a bit embarrassed
Ok to have some technical debt, rather than using perfection as excuse to delay
How many would have guessed that the Middle East is really a hotspot for genomics?
Several countries, numerous projects, funding, and seriousness to implement well
A lot of what we’re seeing in situ spatial transcriptomics is “reeducation” on performance specs. This is because
#CosMx
got to market first, spat out weird benchmarks, and did an incredible marketing campaign.
“Patient” Journey after getting wellness test consisting of ACMG genes, pharmacogenomics, and PRS 🧵
Genetic counselor (GC): hi 👋 thanks for waiting for your results!
Patient (P): it wasn’t easy. It took two visits and three months
GC: yes, the core lab we used lost the 1st!
Illumina’s core competency is to smash NGS competition. They have proven time and time again. And as time passes, their moat gets bigger.
What they have been working on in parallel is…
When an engineer/field applications specialist is working on an instrument in your lab, pls be decent and:
- supply them tools and buffers they may have forgotten
- have the lab empty so they can scream “F’ing Box!” in private or while on a phone call to their base
1/
#illumina
previously said to want to drop sequencing prices to support single cell and pop seq projects. Maybe they might actually feel the pressure from
#NIPT
. Here’s why I speculate this:
Unpopular advice for vendors putting in a bid to win a huge, visible project:
1. Over promise, give yourselves time to deliver
Sell the roadmap
These projects will take years to complete anyway
While pleased that the Times are bringing more genomic tests onto the front page, it comes across as their usual anti-Corp agenda. In addition to the post-article comments 3 bits of context are missing . . .
It really gets me down when…
…protein people say that the lack of correlation between RNA and proteins is greater than the measurement uncertainty proteomics platforms have
In my bio, I said that I'm sharing my personal journey through the twists, turns, ups and downs of working in a genomics core facility, but FYI the things that really get me down, I can't post about 😅. I will probably do a tell-all when I retire 😂😂.
1/ Sequencing approaches to
#spatialbiology
can solve the resolution problem by skirting over the fact that they don’t really sequence the whole transcriptome when at high resolution.
Researchers have shown not to notice nor care
Scene~
CEO who just got series A for setting up a rare disease genomics lab talking with his consultant pathologist:
CEO: Wow awesome we’re gonna bring personalized medicine to healthcare!
Pathologist: is that a marketing thing? What do *you*mean by personalized medicine?
Investment pitches should go onto some open website so that we can review the historical claims made.
Would make for great due diligence and fairness to avoid the exaggerators and in some cases outright liars.
There’s a life sciences company listing 2x “golf attendant” roles in their careers page. Description matches title.
I’m sure there’s context, but, actually probably not.
@timrpeterson
I guess you’re being tongue in cheek, but how else were we, at scale, going to identify all the players are in a tissue, followed by what they do in response to something? These discoveries on their own seem meaningless but they generate leads to follow up w other technologies
Posting about
#pharmacogenomocs
generally gets limited engagement. Alas this is also reflective of its disproportionately low adoption. A shame, given its relatively
- low cost barrier
- high analytical validity
- high clinical utility
Held back by EHRs and prescribing docs 😔
The premise of the glass being half full in this article is:
‘recent layoffs "should not impact the company's new product pipeline as a lot of the work has already been completed."’
This is unusual and maybe overly optimistic
Invitae’s legacy will include creating a bigger market for quest and labcorp.
Alas, vision and first to market are not good bets in mainstream pathology. (…That’s where we want genomics to go right?)
“Haven’t you demonstrated it already”
Passing the concept phase isn’t a guarantee that the potential product will complete the product development process. An RnD team doesn’t behave like a University research lab that can publish irreproducible results.
Respect RnD
Illumina sequencing announcements at JPM through the years, cherry picked for twitter and to distort context 🐄
2009: paired ends “you asked, we delivered!”
2010: HiSeq 2000 “Bam, forget SOLiD!”
2011: MiSeq “democratization, and forget Ion Torrent!”
The wow years