alex rubinsteyn
@iskander
Followers
5,051
Following
3,938
Media
741
Statuses
19,058
Genomics + immunology + ML = personalized cancer immunotherapy. @CompMedUNC @UNC_Lineberger @OpenVax
Durham, NC
Joined June 2007
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
baekhyun
• 522510 Tweets
Linkin Park
• 164434 Tweets
Adalet
• 154267 Tweets
For Asia
• 136829 Tweets
Chester
• 85236 Tweets
小泉進次郎
• 68536 Tweets
ベイマックス
• 40194 Tweets
知的レベル
• 34158 Tweets
エイリアン
• 33584 Tweets
資さんうどん
• 32424 Tweets
雇用統計
• 32177 Tweets
すかいらーく
• 28901 Tweets
大阪府警
• 26338 Tweets
#INZM_HyperBandLive
• 25363 Tweets
#Mステ
• 22430 Tweets
#Number_1
• 21880 Tweets
#ビリスク
• 16876 Tweets
押し出し
• 14361 Tweets
Hyper Band Live M/V
• 13810 Tweets
神ちゃん
• 13591 Tweets
佐藤勝利
• 11231 Tweets
Each wave hits younger and healthier people first before climbing the age brackets to their more vulnerable contacts. Each time we get a few weeks of "if it's so bad, why aren't hospitals full yet?" - right up until hospitals fill up.
It's like Groundhog's Day on Covid Twitter
A widely-shared quote from a South African doctor, saying Omicron causes only mild symptoms, is being taken out of context. She was referring to a small group of young, healthy people and warned of severe disease in other groups
245
2K
4K
17
219
872
Can we fund pre-clinical cancer immunotherapy research by crowdfunding off Twitter?
Let's try!
Context:
@BenjaminGVincen
& I merged our UNC labs & are working on optimizing personalized cancer immunotherapies; ultimate goal: start better phase I trials.
Proposal thread:
1/
6
82
617
Unsurprisingly, RNA extraction method has a big impact on expression analysis
(PC1 = heat stressed vs. unstressed, PC2 = RNeasy vs trizol extraction)
8
113
624
Finally submitted "SARS-CoV-2 peptide vaccine elicits T-cell responses in mice but does not protect against infection or disease" to biorxiv, I'll post a tutorial thread once it's online.
10
46
373
My mom talks with her childhood friend in Kyiv every day. Her friend's family left Kiev to a small village, which they thought would be safer. The village got conquered by Russians and the residents were told they would be evacuated by Red Cross buses. 1/2
9
78
274
So...is MALAT1 even a real thing?
(it's so consistently the top hit in any expression analysis, I'm curious if any studies actually isolate its functionality rather than its pervasive correlation with *everything*)
35
23
221
Research crowd-funding experiment continues!
tl;dr Personalized cancer vaccines don't currently work, I want to fix that by testing (1) better vaccine types (2) better mutation sources from long-read sequencing.
GoFundMe:
Details:
7
55
202
Let's say you want to publish in a top-tier journal and need to have a high accuracy predictor of something of great medical importance, such as survival of cancer patients in response to immunotherapy. The easiest route is just to cheat: (X,y).predict(X)
2
37
209
The Red Cross buses never came and a neighboring family tried to drive back to Kyiv. Their van was shot up by Russian soldiers, killing the mom and several children. Way to Kyiv barred, no food left in the village.
7
61
196
Another N501Y/E484K lineage of SARS-CoV-2 (B.1.1.248) evolved independently, with a different K417 variant (K417T vs. K417N in B.1.351)
Watching the parallel emergence of the same variant combinations at the same time is kinda wild.
11
82
194
Starting to put slides & links online for my class:
Critical readings in biomedical statistics and machine learning (a.k.a. Defense Against the Dark Arts)
10
32
187
Asking in earnest ignorance: have any interesting biological insights emerged from single cell sequencing?
38
17
175
@JessicaValenti
@neva9257
What am I even reading? This is just so...weird. What is "enforced monogamy"? Like, state assigned sexual relationships?
17
5
139
@binarybits
I am extremely concerned about the burden of "don't get hit by cars" getting shifted even further onto pedestrians with "you should have been carrying your safety beacon". I think there is a likely future where AVs decrease deaths of passengers but increase those of pedestrians
3
1
166
Finally have time to write a Deep Learning for Biology syllabus for next year.
Crowdsourcing: which models should I cover? AlphaFold, RFdiffusion, ESM-2, ...?
(what should I add from the nucleic acid world?)
18
27
157
I got a D- in undergrad linear algebra, it just seemed like extremely tedious accounting exercises.
Totally fell in love with graduate linear algebra -- it has lovely theory, ubiquitous use / hugely varied applications (ML, static analysis, complexity), neat generalizations, &c
@TaliaRinger
I've seen so many students, myself included, that feel that some material is just not for them (e.g. Harmonic/Fourier analysis for me), and then later love this material when they come at it with a different angle or motivation.
6
9
210
7
6
146
Surprised at how nonchalantly we're collectively taking SARS-CoV-2 switching up its tropism and cellular entry route.
I guess let's hope that's a one-off trick and not a broader pattern of flexibility?
7
16
137
I *strongly* feel that you shouldn't be able to publish a bioinformatics tool unless it has a continuous integration service which is able to download the source, install its dependencies, & run the tool on example data.
Tests are nice but most of these things don't even run.
16
15
125
@PrancingPapio
It's been my lifelong goal to publish a deeply flawed applied machine learning paper in NEJM.
As a child, I would often dream of leaking test labels into a training set so I could be crowned king of an obscure biomedical community.
0
0
107
@victorlinw
@SystemsVirology
Syrian hamsters have been a workhorse model animal for SARS-CoV-2 infection, it was a big breakthrough last year when it was discovered that they can be infected with wildtype virus and experience disease that mimics what we see in people. Not perfect but better than K18 mice
3
6
103
To minimize "prolonged contact" between unvaccinated children my niece's kindergarten makes the kids switch play groups every ten minutes.
29
11
99
Update: Everyone in the village starving, buses finally came yesterday but not enough for all the people. My mom's friend's family decided to take their chances and caravan behind the buses. Cars stopped by Russian soldiers before they could catch the buses. 1/n
My mom talks with her childhood friend in Kyiv every day. Her friend's family left Kiev to a small village, which they thought would be safer. The village got conquered by Russians and the residents were told they would be evacuated by Red Cross buses. 1/2
9
78
274
2
32
95
@NickytaLeb
OK, I really need to get our T-cell vaccine BSL3 results out in a preprint
(tl;dr we got strong T-cell responses and no protection against SARS-CoV-2-MA in mice)
8
25
90
I've talked for a while about doing a Defense Against the Dark Arts course which surveys high profile biomedical papers whose core claims are false due to abuse of stats/ML.
I'm trying to wrap up the syllabus to submit for next year; taking suggestions for papers you deeply hate
8
16
89
What a weird turn to see John Ioannidis pushing one of sloppiest studies in the deluge of Covid-19 papers. If he weren't an author I would expect the Stanford serostudy to show up in one of his talks as a particularly potent cocktail of bad research practices.
12
17
88
@richardfuisz
This is very SNV centric
I suspect long reads and improved visibility of indels and SVs will dislodge this paradigm of thinking
1
4
86
Sitting in on
@colinraffel
's Deep Learning with Limited Labeled Data seminar ().
Is there a word for the aesthetics of thought? ML concepts (and classes) have a nice...feeling?
2
25
86
@DrRachaelF
Any idea why people don't get rabies vaccines?
Like, I get that the incidence is low but dang the consequences are high
38
3
87
@highscicomm
This is academic misconduct misconduct and I wish you felt safe taking it up with the journal. Everyone who contributes requires either authorship or acknowledgement. Otherwise the senior author is obscuring how the paper got made.
0
0
83
Neon's MHC-II paper is finally out!
2
29
85
@NewRootsinME
@pbleic
The CDC's new role in only setting rules for institutions and businesses without the backing of public trust is really dangerous and breeds resentment/mistrust that will spill over way beyond Covid.
0
7
84
@pbleic
Man...not even an update for Omicron?
I really don't see how the CDC recovers from this, I don't think I know anyone who still pays attention to their guidelines or takes them at face value. People have chosen their own rules and systems, whether more or less cautious.
2
1
83
Good news: BSL3 experiment results came back!
Bad news: ...no difference between groups.
And thus ends my weird year of working on a SARS-CoV-2 peptide vaccine
Might have some consequences for both
@RADVACproject
and maybe in general the idea that T-cells are protective
8
5
80
@mugecevik
@farid__jalali
I think something that is fueling conspiracy-minded thinking is the apparent lack of interest in what appears like the most parsimonious hypothesis: liver injury resulting from previous SARS-CoV-2 infection (by the virus directly or my MIS-C-like auto-immunity). Why exclude it?
4
7
79
UNC is dropping indoor masking with high levels of community Covid infection. My father-in-law is sick so I'm going to embrace socially uncomfortable eccentricity and wear a full face respirator to get in and out of my office.
10
6
78
@erikphoel
Same. I did poorly in HS and undergrad and then had a "I want to be an academic" epiphany in my early 20s.
Without the GRE and subject GRE tests I don't see any path for someone in my situation to get into a decent grad school.
2
6
75
Today is my grandfather Yakov's birthday. After he died, we found some of his father's desperate letters from the last few days before Nazis overran Kiev and executed all the Jews they found.
I'm working on digitizing and translating them here:
6
10
76
New paper on
@biorxivpreprint
:
"Landscape and Selection of Vaccine Epitopes in SARS-CoV-2"
()
Collaboration with
@ChristofCSmith1
@bgvincentlab
@JuliaKodysh
@timodonnell
@glycam
@Olivercgrant
@PEPperCHIP
@erikgarrison
& others.
Thread:
(1/)
9
31
66
I suspect the patchy sleep will catch up to me but so far this has been really fun.
14
0
70
And yes: it's hard in science academia to pay anywhere near biotech salaries, much less those in tech.
The NIH and university pay brackets both assume and enforce cheap labor. Let's treat that as a problem to solve rather than embrace justifications for the situation.
3
14
68
Reading the
@huggingface
guide to quantized LLMs, bitsandbytes & 8-bit matrix multiplication:
0
18
65
I know tons of responsible, rule following, Covid cautious folks who are flying for Christmas gatherings at a time that they know will probably be a massive wave of Omicron.
However people project Covid onto identity, we have no social or political will to avoid mass infection
6
8
65
@OldNewDad
@aliostad
The soldiers in videos have mostly looked like conscripts, they don't have a choice whether to be in the army and likely don't have much training or even very good equipment
4
3
63
IGV view of U87 (glioma cell line) WGS from my rotation student Andy Lee.
Top: Illumina
Middle: PacBio
Bottom: ONT
Large indels in the SHC2 gene visible from PacBio data (del, 522bp del, 228bp del, 3714bp ins). Same variants messy w/ ONT, totally missing from Illumina.
6
12
65
2
0
61
As might be obvious from my tweets about (1) reflecting on work and (2) moving from NY to NC -- I got a new job!
I'm starting a lab in
@UNC_Lineberger
/
@CompMedUNC
/ UNC Genetics.
If you're at UNC and want to meet up, let me know.
7
3
61
Man. Scientific publishing.
Four months to get reviews, submit major revision, two months re-review: two reviewers satisfied, one has pages of additional requests/comments.
I don't know if this process adds enough value relative to how depressing it makes writing a paper.
8
2
59
Like, this is a generational financial crime. Many layers of theft and lying resulting in financial devastation...this coverage just doesn't make sense. Every article I've read seems to be stuck in some kind of reality distortion field.
2
2
58
@VikiLovesFACS
@beverleyturner
@ClareCraigPath
@RCObsGyn
@NHSuk
I really admire your persistence in the face of sometimes hostile and stubborn commitment to imagined counterfactual narratives.
5
0
56
Sent out 20 more sky site invite codes: it's still new/rough but gaining momentum as the spot for science conversations. If you participate in research & want an invite code, let me know!
#immunology
#bcell
#tcell
#genomics
#genetics
#oncology
#deeplearning
#llm
#synbio
#techbio
45
20
57
Starting to brainstorm a "deep sequence models for biologists" class. The plan is to spend a month on intro to ML with PyTorch and then move on to specific useful models / use cases.
Other than AlphaFold and/or ESM what other models would you recommend covering?
13
6
55
It's been a preprint for a while but a much improved version of our paper is out in
@GenomeMedicine
"Landscape and selection of vaccine epitopes in SARS-CoV-2"
With
@ChristofCSmith1
@KOimmuno
@BenjaminGVincen
+ many others
Tweetorial:
1/
2
20
53
Cool immunological phenomenon: some fish have *two* distinct B2M genes and they induce two distinct binding repertoires in the same MHC molecules
The Mechanism of b2m Molecule-Induced
Changes in the Peptide Presentation Profile in a
Bony Fish
3
5
54
Hearing a father and ~8-10 year old son on the street in Brooklyn talk about viral immunity...at a research level. T cells, NK cells, diversity of receptors. Kid just said "lysosome" and I feel like I'm in the twilight zone.
6
1
53
Hadn't seen this before:
Identification of bacteria-derived HLA-bound peptides in melanoma
Curious what you all think of it
3
11
52
Very cool assay: iTAST
"Direct measurement of T cell receptor affinity and sequence from naïve antiviral T cells"
2
13
52
MLCB (Machine Learning in Computational Biology) totally unlinked from NeurIPS:
Now a two day virtual conference November 23 & 24, 9am-5pm PST.
0
23
49
Mild headache ~30 minutes after first NVX-CoV2373 trial injection, feeling hopeful.
4
2
49
Preprint live:
tl;dr We did not disprove the role of T-cells in clearing SARS-CoV-2 (sorry to disappoint) but we did manage to make a vaccine which induces strong but useless T-cell responses against SARS-CoV-2. Details & guess at interpretation below
1/n
Finally submitted "SARS-CoV-2 peptide vaccine elicits T-cell responses in mice but does not protect against infection or disease" to biorxiv, I'll post a tutorial thread once it's online.
10
46
373
1
17
49
I am ridiculously excited to do long read cancer sequencing, especially with the new higher accuracy R10 flow cell. I feel like my whole genomics career has been stuck under a short read streetlight.
Hypothesis: ~0% of neoantigens are SNVs
2
9
48
Rare instance of bioinformatics influencing NLP (instead of the other way around)
Our secret weapon is to prune the output space using an indexing structure called an FM-index (). Using it, we find all continuations in a corpus while constrained decoding. All decoded sequences are then grounded into the document(s) they appear in.
6/n
1
0
13
2
8
50
Do you work with immunological data that requires harmonizing MHC names across multiple species, formats, or sources?
Check out the Python MHC parsing library I've been working on: mhcgnomes ()
Let me know if you find any errors or need other features
3
14
49
@MicrobiomDigest
Every claim of SARS-CoV-2 outside China before January needs to come with a genome sequence. Otherwise it's sewing misinformation and wasting the time of everyone who has to debunk it.
3
6
45
I've been away from science Twitter working on two major projects:
Primarily (1) keeping a pair of infants alive and happy and (2) starting up the Personalized Immunotherapy Research Lab with Ben Vincent and Jared Weiss at UNC.
Get ready to hear a lot more about both!
4
2
47
I have an unfilled post-doc position in computational immunology, if anyone in Ukraine is interested, let me know.
0
13
47
So what actually pops out when you try to explain every read across RNA-seq datasets?
Nice figure from Mangul 2018, good to keep in mind when reading other papers that try to make sense of unmapped reads (eg Poore 2020, which mistakes a lot of the other bins as microbial)
Love this (now older) paper:
ROP: dumpster diving in RNA-sequencing to find the source of 1 trillion reads across diverse adult human tissues
0
2
20
5
9
47
Napkin math on why your percentile rank thresholds for CD8+ T-cell epitope prediction are comically lax.
30mb exome -> let's say half is CDS -> 5M aa possible proteome.
Cut those up into 8-11mers -> 20M peptides
x 6 distinct MHC alleles -> 120M distinct possible pMHCs
(cont)
6
7
47
Publication in Cell/Nature/Science journals is primarily a ritualistic expression of elite cultic practices in the construction of certain styles of analyses and subfigures.
Impossible to distinguish aesthetics from content
2
3
47
It's hard to express how tremendously depressing this week is. Like, I feel it as a dull pain in my chest all the time.
6
3
43
Example from a really elegant mouse study by
@VirusesImmunity
's lab
"We therefore hypothesized that lung CD8+ TRM memory cells may play a more important role in protection of convalescent mice than in vaccinated mice"
5
11
44
I need a good reaction GIF for the feeling I get as I keep stumbling across this paper in literature searches:
"A SARS-like cluster of circulating bat coronaviruses shows potential for human emergence" (2015)
3
13
41
This article:
Long quotes without context or pushback, reader would probably not understand that 8b of deposits got stolen & then there was an insider 600m hack.
+ another that's sympathetic to the sketchy company psychiatrist who prescribed stimulants.
2
6
41
PSA: CS PhDs from NYU typically started at $160k+/year from Google in *2013*.
If you have specialized training and think of a six figure salary in a major city as wealthy, you're being brainwashed by academic culture to undervalue your skills.
Anyone who has been a young PI starting a new lab with postdoc hires that are coming from a wide variety of backgrounds with limited experience in YOUR field knows EXACTLY why they are called “trainees” and why they are not payed six figure salaries.
92
10
91
1
5
42
@JeromeAdamsMD
Avoiding Covid for the past two years has partly depended on ignoring or exceeding CDC guidelines.
I'm sure they're under a lot of political pressure and considering complex social and logistical challenges but the compromises they make should not be disguised as "the science"
3
2
41
Yes! Post-doc opening for my new lab posted:
Focus will be Immunology x Machine Learning for designing better personalized cancer vaccines.
If you're interested in working on research together drop me a line: alex.rubinsteyn[at]
0
30
41
Eyyyyy the
@parkerici
TESLA paper is out:
"Key Parameters of Tumor Epitope Immunogenicity Revealed Through a Consortium Approach Improve Neoantigen Prediction"
1
13
41
It's a niche corner but some really nice research on adaptive immunity comes out of a few institutions in Russia eg
This software that lots of people use to identify T-cell receptors:
And this database of their sequences:
0
8
41
@DSA_Intl_Comm
Never supporting, endorsing, or in any way engaging with a DSA backed candidate again. I suspect anyone else with Soviet roots will feel similarly. You guys just don't get imperialism or know how to actually stand against it, a really sad & fundamental misunderstanding.
0
2
38
History of science Q: it seems like ventilation / air quality eventually became a somewhat taboo topic in public health. When did this happen and why?
One of the many indoor air data from ‘The Carbonic Acid, Organic Matter, and Micro-organisms in Air of Dwellings and Schools.’ (1886)
I wonder what these 19th century scientists would say if they heard that ventilation is not investigated during a 21st century airborne pandemic.
5
42
112
5
11
40
I just don't think that if we restarted this whole system from scratch we would design it so that highly trained scientists in their 30s would be unable to afford home ownership or childcare
0
5
40
I don't think people appreciate the degree to which George Church's lab is a startup incubator / biotech empire.
2
4
37
T cell antigen discoveries reveal a marginal contribution of proteasome splicing to the immunogenic MHC class I antigen pool
Maybe obvious but: "spliced peptides will not...expand the number of targetable antigens for immunotherapies."
1
7
36
Throw away your covariates!
You can guarantee infinitesimally small p-values in predicting survival curve separation (ie which patients respond to a therapy) by selection of random binary features (vs. full dataset labels). Best if 90%+ sparse and n_features/n_samples > 10.
3
1
36
Oldie but ubiquitously influential in T-cell immunology:
T-cell epitope mapping by flow cytometry (1998):
1
8
35
You all have to like this logo because I don't have any more time to email with the designer.
2
0
36
Two years ago...
I started following
#ncov2019
a little after, by surreal coincidence while reading Station Eleven
My friends thought I had become an unhinged prepper when I told them we might have a multi-month pandemic; didn't imagine it would still be going in 2022
2
2
34
It's still a work in progress, but I just made a website for the new
@openvax
lab at Mount Sinai: .
Focus: machine learning applied to immunology, neoantigens, personalized cancer vaccines.
3
9
34