Shubhra Agrawal
@shubhra-a.bsky.social
Techno-commercial role @ Elucidata
FAIR data enthusiast, interested in bioinformatics and AI
Personal account
FAIR data enthusiast, interested in bioinformatics and AI
Personal account
Reposted by Shubhra Agrawal
Excited to launch our AlphaGenome API goo.gle/3ZPUeFX along with the preprint goo.gle/45AkUyc describing and evaluating our latest DNA sequence model powering the API. Looking forward to seeing how scientists use it! @googledeepmind
June 25, 2025 at 2:29 PM
Excited to launch our AlphaGenome API goo.gle/3ZPUeFX along with the preprint goo.gle/45AkUyc describing and evaluating our latest DNA sequence model powering the API. Looking forward to seeing how scientists use it! @googledeepmind
Reposted by Shubhra Agrawal
Saying LLMs are worthless/pernicious because they sometimes make mistakes & is causing a general decline in reading skills (true) is like saying Google Maps is worthless/pernicous just because it sometimes gives you bad directions & has caused a general decline in human navigation skills (also true)
June 17, 2025 at 11:20 AM
Saying LLMs are worthless/pernicious because they sometimes make mistakes & is causing a general decline in reading skills (true) is like saying Google Maps is worthless/pernicous just because it sometimes gives you bad directions & has caused a general decline in human navigation skills (also true)
Reposted by Shubhra Agrawal
Eli Lilly nears $1.3 billion deal for gene-editing biotech Verve Therapeutics, FT reports reut.rs/3Zy8t2b
Eli Lilly nears $1.3 billion deal for gene-editing biotech Verve Therapeutics, FT reports
Eli Lilly is in advanced talks to acquire gene editing startup Verve Therapeutics for up to $1.3 billion, the Financial Times reported on Monday, citing people familiar with the matter.
reut.rs
June 17, 2025 at 12:00 AM
Eli Lilly nears $1.3 billion deal for gene-editing biotech Verve Therapeutics, FT reports reut.rs/3Zy8t2b
AI agents fail when built on fragmented, context-agnostic data. In discovery, translational and clinical, that failure costs time, capital, and patient outcomes.
On June 20, we are hosting an in-person workshop in Boston on Building AI Agents with Fit-for-Purpose Data.
On June 20, we are hosting an in-person workshop in Boston on Building AI Agents with Fit-for-Purpose Data.
June 16, 2025 at 10:29 PM
TIL 10/365: blood genotyping is not as straightforward as most of us have been taught. the most surprising part for me was that some people can safely take blood from 1 type but can't donate to the same type.
this TIL is inspired by the thread here -> bsky.app/profile/alan...
this TIL is inspired by the thread here -> bsky.app/profile/alan...
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 21, 2025 at 9:08 PM
TIL 10/365: blood genotyping is not as straightforward as most of us have been taught. the most surprising part for me was that some people can safely take blood from 1 type but can't donate to the same type.
this TIL is inspired by the thread here -> bsky.app/profile/alan...
this TIL is inspired by the thread here -> bsky.app/profile/alan...
TIL 9/365: Boehringer Ingelheim has an open science portal opnMe where they pose unsolved challenges in different therapeutics areas and invite participants to share their solutions. Promising ideas get funding support!
Check out the list of challenges: www.opnme.com/opn2experts
Check out the list of challenges: www.opnme.com/opn2experts
January 21, 2025 at 8:35 PM
TIL 9/365: Boehringer Ingelheim has an open science portal opnMe where they pose unsolved challenges in different therapeutics areas and invite participants to share their solutions. Promising ideas get funding support!
Check out the list of challenges: www.opnme.com/opn2experts
Check out the list of challenges: www.opnme.com/opn2experts
TIL 8/365: Phone SIM cards have a shelf life 😆
Sounds obvious in retrospect, never considered it as a possible explanation for my near constant network failure. ran around customer care portals for a month before someone suggested i change it 🫠
Sounds obvious in retrospect, never considered it as a possible explanation for my near constant network failure. ran around customer care portals for a month before someone suggested i change it 🫠
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 21, 2025 at 8:27 PM
TIL 8/365: Phone SIM cards have a shelf life 😆
Sounds obvious in retrospect, never considered it as a possible explanation for my near constant network failure. ran around customer care portals for a month before someone suggested i change it 🫠
Sounds obvious in retrospect, never considered it as a possible explanation for my near constant network failure. ran around customer care portals for a month before someone suggested i change it 🫠
TIL 7/365: Being a 'woman' is hard
i decided to learn the basics of styling and makeup this year. but beyond the active time spent shopping/trying products on, there is so much homework that the average woman does on fabrics,design,skin types,hair types,markets and trends.
i feel like a toddler 🧒
i decided to learn the basics of styling and makeup this year. but beyond the active time spent shopping/trying products on, there is so much homework that the average woman does on fabrics,design,skin types,hair types,markets and trends.
i feel like a toddler 🧒
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 21, 2025 at 8:17 PM
TIL 7/365: Being a 'woman' is hard
i decided to learn the basics of styling and makeup this year. but beyond the active time spent shopping/trying products on, there is so much homework that the average woman does on fabrics,design,skin types,hair types,markets and trends.
i feel like a toddler 🧒
i decided to learn the basics of styling and makeup this year. but beyond the active time spent shopping/trying products on, there is so much homework that the average woman does on fabrics,design,skin types,hair types,markets and trends.
i feel like a toddler 🧒
Reposted by Shubhra Agrawal
Here goes—-on with my day!
January 21, 2025 at 3:39 PM
Here goes—-on with my day!
TIL 6/365: US is now going to control the export of cell sorters and mass spec machines because biological data + AI poses a security risk in the wrong hands.
huh. i'm not familiar with wet-lab, unfortunately. it might make large scale NGS harder in the short term but i'm sure there are ways ??
huh. i'm not familiar with wet-lab, unfortunately. it might make large scale NGS harder in the short term but i'm sure there are ways ??
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 17, 2025 at 4:48 PM
TIL 6/365: US is now going to control the export of cell sorters and mass spec machines because biological data + AI poses a security risk in the wrong hands.
huh. i'm not familiar with wet-lab, unfortunately. it might make large scale NGS harder in the short term but i'm sure there are ways ??
huh. i'm not familiar with wet-lab, unfortunately. it might make large scale NGS harder in the short term but i'm sure there are ways ??
Reposted by Shubhra Agrawal
Meet gander, a chat interface for R data scientists🪿
gander automatically incorporates context from your #rstats environment and surrounding code and can be used with any model supported by ellmer in either RStudio or Positron.
Read more: simonpcouch.github.io/gander/
gander automatically incorporates context from your #rstats environment and surrounding code and can be used with any model supported by ellmer in either RStudio or Positron.
Read more: simonpcouch.github.io/gander/
January 16, 2025 at 2:47 PM
Meet gander, a chat interface for R data scientists🪿
gander automatically incorporates context from your #rstats environment and surrounding code and can be used with any model supported by ellmer in either RStudio or Positron.
Read more: simonpcouch.github.io/gander/
gander automatically incorporates context from your #rstats environment and surrounding code and can be used with any model supported by ellmer in either RStudio or Positron.
Read more: simonpcouch.github.io/gander/
TIL 5/365: My ADHD ass is dysregulated 90% of the time. I just never realised it before because I associate emotional dysregulation with bursts of temper, or extreme mood swings (:
apparently, flip-flopping between internal screaming and static-noise-no-thought phases counts too.
apparently, flip-flopping between internal screaming and static-noise-no-thought phases counts too.
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 17, 2025 at 9:04 AM
TIL 5/365: My ADHD ass is dysregulated 90% of the time. I just never realised it before because I associate emotional dysregulation with bursts of temper, or extreme mood swings (:
apparently, flip-flopping between internal screaming and static-noise-no-thought phases counts too.
apparently, flip-flopping between internal screaming and static-noise-no-thought phases counts too.
TIL 4/365: I am helpless against the Lego shelves 😩
Especially the 3-in-1 kind
Especially the 3-in-1 kind
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 15, 2025 at 8:21 PM
TIL 4/365: I am helpless against the Lego shelves 😩
Especially the 3-in-1 kind
Especially the 3-in-1 kind
Reposted by Shubhra Agrawal
Please for the love of Satan, post more jokes!!
btw if we resolve to keep our individual joke post:the-situation-is-dire post ratio at 3:1 min (pref 5:1 or better) we could save this site from becoming twitter. “Becoming twitter” is a fate worse than death so maybe consider this concept. In order to lead my example I will next post about butts
January 14, 2025 at 2:55 PM
Please for the love of Satan, post more jokes!!
TIL 3/365: there is an upper limit to how many people you can directly manage while working on active projects yourself :|
i also think I blew wayy past this limit months ago. any sage advice from people who've been there? 😅
i also think I blew wayy past this limit months ago. any sage advice from people who've been there? 😅
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 14, 2025 at 6:36 PM
TIL 3/365: there is an upper limit to how many people you can directly manage while working on active projects yourself :|
i also think I blew wayy past this limit months ago. any sage advice from people who've been there? 😅
i also think I blew wayy past this limit months ago. any sage advice from people who've been there? 😅
Reposted by Shubhra Agrawal
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 12, 2025 at 2:18 PM
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
Reposted by Shubhra Agrawal
With 100 millions single cell gene expression measurements, using >1000 drugs across 60 cell line models, Tahoe-100M is larger than all public single-cell transcriptomic data combined. And it is 50x larger than all drug-perturbed single-cell data put together.
January 13, 2025 at 4:23 PM
With 100 millions single cell gene expression measurements, using >1000 drugs across 60 cell line models, Tahoe-100M is larger than all public single-cell transcriptomic data combined. And it is 50x larger than all drug-perturbed single-cell data put together.
TIL 2/365: core facilities in academia have a transparency problem
researchers get access to the raw data, and the processed excel file from the core. The in-between details important for drawing any verifiable conclusions? gone.
even basics like the target library used for LCMS could be missing 🤷♀️
researchers get access to the raw data, and the processed excel file from the core. The in-between details important for drawing any verifiable conclusions? gone.
even basics like the target library used for LCMS could be missing 🤷♀️
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 13, 2025 at 6:29 PM
TIL 2/365: core facilities in academia have a transparency problem
researchers get access to the raw data, and the processed excel file from the core. The in-between details important for drawing any verifiable conclusions? gone.
even basics like the target library used for LCMS could be missing 🤷♀️
researchers get access to the raw data, and the processed excel file from the core. The in-between details important for drawing any verifiable conclusions? gone.
even basics like the target library used for LCMS could be missing 🤷♀️
TIL 1/365
lncRNAs might be a lot more important for single cell: cell type annotation than I previously thought (okay, I never thought about it before 🤷♀️)
most of the current public SC datasets don't even capture lncRNAs consistently. def going to explore this more
lncRNAs might be a lot more important for single cell: cell type annotation than I previously thought (okay, I never thought about it before 🤷♀️)
most of the current public SC datasets don't even capture lncRNAs consistently. def going to explore this more
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 12, 2025 at 2:46 PM
TIL 1/365
lncRNAs might be a lot more important for single cell: cell type annotation than I previously thought (okay, I never thought about it before 🤷♀️)
most of the current public SC datasets don't even capture lncRNAs consistently. def going to explore this more
lncRNAs might be a lot more important for single cell: cell type annotation than I previously thought (okay, I never thought about it before 🤷♀️)
most of the current public SC datasets don't even capture lncRNAs consistently. def going to explore this more
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
a man with his mouth open and the words one after the other behind him
Alt: gif of David from Schitts creek saying "one after the other, consecutively, in a row"
media.tenor.com
January 12, 2025 at 2:18 PM
One of my focus areas for 2025 is to 'build publicly' at work. Not a novel concept, but definitely an area of improvement for me.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
but why stop there? so i'll be starting a TIL (Today, I learned) series on bsky with the aim to share 365 TIL posts this year.
Reposted by Shubhra Agrawal
Quarto is great! Another option that elegantly addresses this issue is jupytext: jupytext.readthedocs.io/en/latest/, which creates a version-control friendly twin of your notebook and keeps them in sync as you work.
Jupytext — Jupytext documentation
jupytext.readthedocs.io
January 10, 2025 at 5:07 PM
Quarto is great! Another option that elegantly addresses this issue is jupytext: jupytext.readthedocs.io/en/latest/, which creates a version-control friendly twin of your notebook and keeps them in sync as you work.
Reposted by Shubhra Agrawal
How to build a virtual cell is the “Holy Grail” of the biology! A great read from The Atlantic @theatlantic.com ! Also thanks for highlighting our scGPT, the model that started the new wave of foundation models in biology!
www.theatlantic.com/technology/a...
www.theatlantic.com/technology/a...
A Virtual Cell Is a ‘Holy Grail’ of Science. It’s Getting Closer.
Large language models may unlock a new and valuable type of research.
www.theatlantic.com
January 10, 2025 at 7:37 PM
How to build a virtual cell is the “Holy Grail” of the biology! A great read from The Atlantic @theatlantic.com ! Also thanks for highlighting our scGPT, the model that started the new wave of foundation models in biology!
www.theatlantic.com/technology/a...
www.theatlantic.com/technology/a...
Reposted by Shubhra Agrawal
When asked about knockout vs knockdown, Google AI wonderfully bet-hedges on whether you're interested in genetics or boxing
January 10, 2025 at 7:44 PM
When asked about knockout vs knockdown, Google AI wonderfully bet-hedges on whether you're interested in genetics or boxing
Reposted by Shubhra Agrawal
Btw - this is another case where my frequent point that not just IF but also PubMed/PMC are obstacles to innovation applies: if a journal does a novel form of post-hoc change that doesn't fit in PubMed's rigid framework, it's not registered (so is disincentivized)
There's a broader point here about even legal proliferation of copies of articles in multiple locations. Given no obligation (or tech or $ or willingness..) for a site to update these, copies may not reflect the current version. That may suit bad actors, plus contextual misinformation is a risk 1/2
84.83% of retracted articles available via Sci-Hub do not mention their retraction status
www.tandfonline.com/doi/full/10....
www.tandfonline.com/doi/full/10....
January 7, 2025 at 3:31 PM
Btw - this is another case where my frequent point that not just IF but also PubMed/PMC are obstacles to innovation applies: if a journal does a novel form of post-hoc change that doesn't fit in PubMed's rigid framework, it's not registered (so is disincentivized)
Reposted by Shubhra Agrawal
Part of the reason people are glum about AI is that researchers themselves represent the goal as full equivalence for human agency (“AGI”).
While I get the theoretical interest, I think what most customers want (and what we’ve really produced so far) is closer to “prosthetics for thinking.”
While I get the theoretical interest, I think what most customers want (and what we’ve really produced so far) is closer to “prosthetics for thinking.”
January 5, 2025 at 4:07 PM
Part of the reason people are glum about AI is that researchers themselves represent the goal as full equivalence for human agency (“AGI”).
While I get the theoretical interest, I think what most customers want (and what we’ve really produced so far) is closer to “prosthetics for thinking.”
While I get the theoretical interest, I think what most customers want (and what we’ve really produced so far) is closer to “prosthetics for thinking.”