Steven Salzberg 💙💛
@StevenSalzberg1 · Maryland, USA · joined 26 Aug 2012
Bloomberg Distinguished Professor of BME, CS, and Biostats at Johns Hopkins Univ., tennis player, https://t.co/xwqVB2TYkL, @stevensalzberg.bsky.social
19 394Followers
258Following
7 404Posts total
214.4KViews on collected posts
Latest posts
Looking for new posts…
700 known cases of measles in Pennsylvania probably means at least 10,000 actual cases. The anti-vax movement (and RFK Jr.) is to blame, as I've written before, e.g. this 2019 article (@forbes, free on Substack): https://t.co/rXJLvMaFO6 https://t.co/iHicH7bPWY
976 views · 6 likes
· 1 reposts · 1 replies
Open on X →
@StevenSalzberg1 Releasing AlphaGenome Atlas is still useful for the community. They never presented AVI as ground truth; it’s a research database for ranking variants and generating hypotheses.
Practically, people were already querying the API for single-variant effects — now t
1.8K views · 15 likes
· 0 reposts · 1 replies
Open on X →
So: 9 billion predictions with almost no proof behind any of them. Highly implausible. Many complex figures that are poorly explained, almost certainly written in part by AI. A lengthy mss that will overwhelm reviewers who will just throw up their hands in surrender 13/14
2K views · 11 likes
· 3 reposts · 1 replies
Open on X →
And of course @GoogleDeepMind issued a press release, and @Nature already wrote a news story. Who wants to bet that @nature publishes this within 6 months? Yet I can find nothing interesting here. 14/14
2K views · 22 likes
· 1 reposts · 1 replies
Open on X →
... look at bases near splice sites. We know those (noncoding) bases have a big impact too! I can write a program and generate scores showing these "variant impacts" in an afternoon–but why would I? ... 12/N
2.9K views · 10 likes
· 1 reposts · 1 replies
Open on X →
...if you want to know if a mutation has an impact, we have many published methods for doing that, dating back decades. E.g., just look in protein-coding regions for changes that affect the amino acid translation. We know those have a big impact already! Then ... 11/N
1.9K views · 5 likes
· 0 reposts · 1 replies
Open on X →
But back to the main result: the 9 billion AVI scores. What the heck does one do with this? They didn't prove that any of them are meaningful (okay, maybe 0.0001% of them), so this is just more AI slop to me, because look... 10/N
2K views · 7 likes
· 0 reposts · 1 replies
Open on X →
... the top-ranked AVI score for one patient was in an intron of a gene called DNM1. Turns out this variant was reported in 2 previously published cases. They did some cell-line expts to show it changes splicing in DNM1. So it's plausible! But there's so much missing here 8/N
2.1K views · 7 likes
· 0 reposts · 1 replies
Open on X →
But how about some actual biology? It's there, but only in an anecdote about how they helped to resolve a rare disease case. Hmm, what about that? Well, first they ranked variants in 814 "unsolved" cases. (No explanation of how these variants were initially collected.) 7/N
2.2K views · 8 likes
· 0 reposts · 1 replies
Open on X →
First, this is obviously cherry-picking a case where they found something plausible. What about the other 813 unsolved cases? Anything at all there? And they don't explain how many variants they started with; it appears that some serious filtering was done up front 9/N
2K views · 10 likes
· 0 reposts · 1 replies
Open on X →
... and even for the mutations that have been observed (a small fraction of the 9 billion), we only have imprecise biological measurements of their impact. So the claim that these AVI scores are useful is wildly implausible on its face. But I read the paper... 5/N
3K views · 17 likes
· 3 reposts · 1 replies
Open on X →
As far as I can tell, the AVI score is some kind of estimate of how "important" a DNA base is. But there are some gargantuan problems here. First, there isn't enough biological data in the universe to train a model on these 9 billion mutations. Most have *never* been observed 4/N
2.9K views · 28 likes
· 3 reposts · 1 replies
Open on X →
... trying to find something that would convince me that at least a subset of the scores might be useful. Instead, I found a enormous number of claims, mostly supported by comparing AlphaGenome to other computational methods, usually with very thin or no explanation 6/N
2.3K views · 13 likes
· 1 reposts · 1 replies
Open on X →
I have some thoughts about this: a thread. "DeepMind’s new genome ‘atlas’ charts effects of all nine billion human gene mutations" https://t.co/L8ow4uvuto TL;DR: AI slop is taking over genomics 1/N
45.7K views · 189 likes
· 40 reposts · 12 replies
Open on X →
... each base can be mutated to 3 other bases. For each of these 9 billion changes, they report its "impact" as something they called AVI, for AlphaGenome Variant Impact." What does this even mean? Well, I've spent a few hours now with the paper and ... 3/N
3.1K views · 7 likes
· 0 reposts · 1 replies
Open on X →
Well, the folks at @GoogleDeepMind have produced another massive paper about their human DNA "foundation model," AlphaGenome. This time, they have produced scores for every possible single-base mutation in our 3-gigabase genome. That's 9 billion predictions because 2/N
3.8K views · 9 likes
· 1 reposts · 1 replies
Open on X →
Check out our new preprint on thousands of novel recursive exons in the human genome (if you don't know about recursive exons, they are pretty cool), led by PhD student David Bass: https://t.co/tMeYogDlEQ
14.2K views · 151 likes
· 28 reposts · 4 replies
Open on X →
to add a bit more explanation: when you are looking at millions of mutations with no hypothesis in mind, you are implicitly considering a mind-boggling huge number of hypothesis. Large-scale genome scans always find 1000s of correlations; that doesn't make any of them real
334 views · 1 likes
· 0 reposts · 2 replies
Open on X →
@StevenSalzberg1 They followed standard GWAS methodology and validated results within family. Do you think that's wrong?
6.2K views · 75 likes
· 0 reposts · 4 replies
Open on X →
I'm sorry, but I don't believe this result is worth the paper it's printed on. Sure, if you look at trillions of combinations and don't correct for that, you'll find associations, purely by chance. But they got the NYT headline, didn't they? https://t.co/MUnjeeDwI4
79.7K views · 83 likes
· 6 reposts · 17 replies
Open on X →
you can't blame excess weight on your gut microbiome, as this extensive new study by @NiranjanTW shows
7K views · 16 likes
· 1 reposts · 2 replies
Open on X →
Are there robust associations between the gut microbiome and obesity?
We explored this extensively with populations-scale metagenomic data for >800 South-East Asians in the HELIOS cohort @macadology
https://t.co/srPOeHB7Qz
8.8K views · 16 likes
· 1 reposts · 1 replies
Open on X →
Very pleased to see the 2023 Warren Alpert Foundation Prize going to David Lipman, co-creator of BLAST and founding director of NCBI, GenBank, PubMed, and more. I'm also happy to say that he's been a good friend and colleague of mine for many years https://t.co/m8tZxCO1o4
17.2K views · 72 likes
· 13 reposts · 1 replies
Open on X →
Against accounts of the same size
22 posts from the last 90 days, next to the 10K–100K follower range. shown widely, but few of those viewers react.
Median views2 628this account924median for 10K–100K
Reach, %13.55%this account3.62%median for 10K–100K
Engagement, %0.54%this account1.52%median for 10K–100K
| Metric | This account | Median for 10K–100K | Ratio |
|---|---|---|---|
| Median views per post | 2 628 | 924 | 2.84× |
| Reach (views ÷ followers) | 13.55% | 3.62% | 3.74× |
| Engagement rate | 0.54% | 1.52% | 0.35× |
Others in this range → Compare with another account → How these benchmarks are built →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
45.7K17 Sep
2.3K
2.9K
3K
2K
2.2K
2.1K
2K
1.9K
2.9K
2K
2K
1.8K
97620 Sep
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.54%17 Sep
0.64%
1.10%
0.70%
0.54%
0.41%
0.37%
0.40%
0.31%
0.41%
1.20%
0.75%
0.88%
0.82%20 Sep
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes61.5%778 in total
Reposts8.1%103 in total
Replies4.6%58 in total
Quotes1.3%16 in total
Bookmarks24.6%311 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
Followers by day
21 Sep
Daily snapshots since 21 Sep 2026; the dashed line is the starting count.
Similar accounts
Miguel Abellán
@miguelabellan
38.8K followers
· 17.7K posts
Alison Bechdel
@AlisonBechdel
38.8K followers
· 662 posts
Supa 🦜
@Supa_LoL
38.8K followers
· 3.4K posts
Amy Aiello Lofgren
@azsweetheart013
38.8K followers
· 38.6K posts
Collecting Cars
@collectingcars_
38.8K followers
· 27.6K posts
Nathan Patterson ✓
@NathanPatterson
38.8K followers
· 164 posts
cami ✓
@camiinthisthang
38.8K followers
· 146.7K posts
Alex Kack
@Alex_Kack
38.8K followers
· 31.1K posts