NimwegenLab
@NimwegenLab
Followers
3,876
Following
198
Media
177
Statuses
4,643
Gene regulatory networks and genome evolution. How do single cells make up their minds? @NimwegenLab @mstdn .science
Basel City, Switzerland
Joined February 2014
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Marcellus Williams
• 465507 Tweets
Missouri
• 267391 Tweets
Vinicius
• 121790 Tweets
Mike Parson
• 91144 Tweets
Alavés
• 65940 Tweets
Ghost of Yotei
• 65811 Tweets
#StateofPlay
• 60680 Tweets
発売日決定
• 56306 Tweets
Endrick
• 53008 Tweets
Horizon
• 49972 Tweets
SCOTUS
• 47995 Tweets
#GHLímite3
• 46978 Tweets
Quito
• 35632 Tweets
Lunar
• 32134 Tweets
アルシュベルド
• 28887 Tweets
メインモンスター
• 28716 Tweets
Colo Colo
• 27272 Tweets
ラヴィット
• 27136 Tweets
Vallejo
• 25766 Tweets
Willie
• 14992 Tweets
アイドル19歳
• 14427 Tweets
Stellar Blade
• 12426 Tweets
保険証廃止
• 11985 Tweets
Guápulo
• 11815 Tweets
Soul Reaver
• 10517 Tweets
I want to state that I agree with most of the article below. This may well turn out to be the biggest scientific scandal of our lifetimes and the silence of us scientists on this may come to be viewed as a huge moral failure in the future.
66
482
1K
I see we are getting to the stage of the discussion where people are starting to defend UMAP saying it can 'reveal patterns' or 'structure' in the data. Without ever specifying what precisely these patterns/structures represent. This is not surprising because hardly anybody 1/n
Thoughtful, well-informed counterpoint on the use of UMAP for visualising human genetic data, following the intense debate about the All Of Us paper this week.
1
5
27
4
129
579
Hey, weren't you a co-author on that?
Yes, I was, but the Nature editors neutered the letter to the point that I felt I could no longer in good conscience put my name on it. The most egregious (and final straw for me) was the insistence to change 1/n
12
136
529
How many distinct gene expression states are present in a given scRNA-seq dataset, and what are they? Maybe surprisingly, as Pascal Grobecker shows in his PhD work, there is a unique rigorous answer to this question, free from any tunable parameters. 1/n
10
79
335
I concur with
@Ayjchan
and wish more colleagues would speak up on. 20 million dead and somehow we're all fine that we're not going to find out where this came from?
15
104
322
Loud sobbing reported from under reverend Bayes' grave stone.
Seriously, I might use this as an example in my class to show how NOT to do statistics. Note that the CI on specificity includes false positive rates larger than the observed fraction of positives.
Preprint: In Santa Clara County CA, ~2.5-4.1% of residents (48,000- 81,000 people) have had
#covid19
, based on antibodies. That’s 50-85x the confirmed cases.
26
158
291
12
86
241
It's so depressing what a f*ing clown show COVID origins reporting has become. The Atlantic, Science, NYT, the Guardian. None are apparently able to do the slightest bit of due diligence on what this (so far unavailable) data would even mean.
26
26
248
Can anybody explain me what the reasonable justification for this is? Because I am starting to become a bit perplexed about how little outcry there seems to be about this in the scientific community.
NIH sent The Intercept 292 fully redacted pages related to virus research in Wuhan by
@fastlerner
17
130
318
12
44
219
Did you know that, in E. coli, the structure of the gene regulatory network does not only control the average expression levels of genes, but also their noise levels, i.e. that expression noise is itself a regulated quantity? Check out our paper. 1/n
4
52
213
When analyzing scRNA-seq data, shouldn't it be possible to rigorously correct for both the Poisson sampling noise that is intrinsic to the process of gene expression, and the sampling noise introduced by the measurement (i.e. mRNA capture and sequencing)? Well, yes! 1/n
7
63
215
This is not how one does serious science. One cannot show a picture, claim it shows the 'structure in the data' and then be completely unable to explain what this 'structure' supposedly represents.
It's a joke, frankly. n/n
6
20
206
A good summary of the evidence that a lab leak is the most parsimonious explanation for the origin of SARS-CoV2.
I know many in the scientific community prefer to just look away, but IMO this is justifiably undermining trust and highly damaging to science.
Ahead of today's hearing in Congress: my opinion piece with the
@nytimes
on why Covid-19 was likely caused by a lab accident.
My hope since 2020 has been for leaders, especially scientists, to lead the charge in investigating a plausible lab
#OriginOfCovid
- as opposed to
461
1K
5K
12
46
176
@jbkinney
is a top quality scientist with the rare spine to speak up on SARS-CoV2 origins and the need for better biosafety. I'm disgusted that he's apparently being threatened for this.
And yes, I'm by now also convinced SARS-CoV2 did most likely originate from a lab accident.
I am leaving
@BiosafetyNow
. I always knew there was some professional risk to me speaking out, but my advocacy has now become a liability that I can no longer ignore. 1/5
103
212
917
1
60
170
I could hardly disagree with this more. If you write a PhD thesis you should be able to formulate in your own words how your research fits into a larger context. In fact, I find these parts of PhD theses often the most informative regarding how scientifically mature and 1/n
About 10 years ago, I was on a PhD committee. It wasn't until the evening before the defence that I made my way through the thesis. The student was not a native English speaker, his grammar was very creative, and it was tough going.
In chapter 2 the English grammar dramatically
272
274
3K
11
15
171
You only know what your measurements say about your system to the extent that understand the noise in your measurements. This applies especially to scRNA-seq data, which is subject to severe sampling noise. See here our attempt at solving this problem: 1/n
1
56
176
Big weight off my shoulders today. I'm very grateful to the eLife editors that agreed to go through a huge pile of reviews, revisions, and rebuttals from a previous journal and come to a decision so rapidly. 1/n
4
48
175
It just keeps getting worse, doesn't it?
Let's pretend we're going to do these GOFROC experiments at BSL-3 in the US but actually do them at BSL-2 in China.
What a great idea.. what's the worst that could happen?
4
47
166
The strange state of current publishing. I think some stats on our recent Cellstates paper provide interesting food for thought. The preprint was put on BioRxiv and within 3 months there were ~8500 abstract views and >2000 PDF downloads.
We also submitted it to PCB 6-11-2023 1/n
5
46
168
Physical implementation of the UMAP algorithm.
0
13
163
Perfect example of what is so terribly wrong with this field. No explanation at all of how it works or why it is better. We know it's mathematically impossible to capture all structure in an arbitrary high-dim dataset in 2D. So Q is: what structure does 'deep TDA' decide to keep?
Another deep learning breakthrough:
Deep TDA, a new algorithm using self-supervised learning, overcomes the limitations of traditional dimensionality reduction algorithms.
t-SNE and UMAP have long been the favorites. Deep TDA might change that forever.
Here are the details:
48
685
4K
8
31
162
@sciencecohen
@jbloom_lab
For me the biggest story is NCBI refusing Jesse's suggestion to look for SARS-CoV2 reads in other retracted submissions which, IMHO, boils down to saying they don't WANT to learn more about the origins of SARS-CoV2. Why not? This seems very troubling.
5
38
142
This might be the longest I've ever worked on one paper! Puzzled by
patterns in our data, we spend years coming to grips with the role of recombination in bacterial genome evolution. But now we finally have tools to get a quantitative handle on this.
4
86
141
I virtually always agree with
@trvrb
's on COVID issues, but here is an exception. We don't have enough information to say with any certainty how SARS-CoV2 originated, but I don't agree at all that the evidence strongly favors zoonosis. Here's my view. 1/n
4
26
107
Jesus. This looks pretty devastating to me for Pekar et al. It's so bad I almost find it hard to believe. Did they really just go and selectively remove genomes that would have invalidated the conclusion that they preferred? Really?
@humblesci
Daoyu
@ydeigin
@quay_dr
and myself have published a study that shows the exclusion of 20 A/B intermediate SARS2 genomes from the analysis of Pekar et al. is unwarranted 🧵
33
81
221
8
35
133
It is extremely depressing that there is even a need for this thread. Apparently some people believe that inconveniencing a few virologists is worse than rolling the dice with killing millions of people.
With all due respect for fact that different people have differing motives and sincerity, I think we need both more investigation of COVID-19 origins and more careful thought about safety of work with potential pandemic pathogens.
30
88
519
3
28
122
Am I the only one that finds these 'types of XXX papers' memes incredibly depressing? It seems to essentially say: Ha ha ha.. look at us.. we pretend to be trying to solve real scientific problems.. but actually we're facetiously churning out cooky-cutter bullshit.. ha ha ha!
14
1
128
Finally found time to read the paper beyond quickly browsing through it. I'm afraid my disbelief has only increased. Apart from specific technical points, if a paper is questioning long and widely held beliefs that appear to be supported by lots of observations, one would 1/n
Does anybody else find this hard to believe also? All those patterns with evolutionary rate and dN/dS rates as function of gene expression and essentiality were all mistaken? I really have a hard time wrapping my head around this at first reading.
13
9
95
5
43
127
title and pointed out there isn't any evidence, the Nature editor defended the title change by saying that there could be evidence IN THE FUTURE... Right.
I note that the unneutered BioRxiv version is still
available at n/n
5
8
126
Very good. Been waiting for a competent and spin-free analysis like this of the sequencing data of the Huanan market samples. And it blows my mind that all this media noise was really only about 1 single read in 1 of 14 samples with abundant raccoon dog DNA.
In new study, I have analyzed correlation between SARS-CoV-2 & animal genetic material in full set of environmental samples from Huanan Seafood Market.
Analysis clarifies what sequencing these samples can & cannot tell us about early outbreak at market.
40
345
952
5
24
121
How to rigorously identify distinguishable gene expression states in scRNA-seq data. Our Cellstates paper has now been published:
See the quoted thread for a description of this work.
What are the main revisions in the final paper? 1/n
How many distinct gene expression states are present in a given scRNA-seq dataset, and what are they? Maybe surprisingly, as Pascal Grobecker shows in his PhD work, there is a unique rigorous answer to this question, free from any tunable parameters. 1/n
10
79
335
3
19
115
Remember: mRNA != protein, and mRNA expression comes in bursts that are much shorter lived than the proteins they produce. I know, for example, that one can FACS sort neural stem cells based on Hes5 expression and find no Hes5 mRNA in a good fraction of the Hes5 positive cells.
Single cell Twitter:
How can it be that popular cell assignment tools assign CD4+ T-cells to lots of T-cells totally lacking CD4?
I know the classics “it’s a dropout” or “heterogeneity”, but still, this lacks biological logic. Don’t overlook basic plots & controls on your data!
17
9
95
4
9
114
Does it get more people to read it? No.
What it gets us is that it is now officially peer-reviewed and published in a good journal. This imprimatur costs $2800 and 10 months with lots of, in my opinion, unecessary headaches. And this is a BEST case scenario. Makes one think. n/n
4
17
113
Oh god this makes me want to cry. What did you expect? That if you keep adding lower signal-to-noise genes to your matrix the fraction of variance captured by 200 PCs would go up?
More genes, more sequencing, more information in your data matrix, that's always better, right?!
According to single-cell current standards, not quite.
In >100 datasets from
@humancellatlas
and beyond, PCA preserves only ~30% of total covariance using defaults (200PC).
17
37
224
1
6
109
There are two types of statisticians: Those that have understood that p-values are answering the wrong question, and those that have not understood this yet. There is no 'disagreement', it is just a failure to grasp the point.
9
9
104
Growth-rate controls the sensitivity of gene regulatory
circuits. Sometimes the most profound insights are those that seem obvious in hindsight. I believe this work of
@ThomasJulou
is one of those. Check it out. And here's a thread to whet your appetite.
2
43
103
No, tSNE/UMAP don't 'preserve neighborhoods'.
Fig 2d shows that tSNE and UMAP get the large majority of k nearest neighbors wrong on all datasets. And this is 'recall' not even of true cell positions but just the logp1 mappings of the raw data!
The authors highlight that UMAP/tSNE preserve neighborhoods and conclude: "If used properly, 2D (tSNE/UMAP) embeddings will continue to help charting the world of single-cell biology."
@hippopedoid
@JanLause
4
16
75
4
21
100
Q for the twitter hive. Has anybody convincingly quantified what fraction of accessible DNA in mammals is bound by DNA binding factors, i.e. is open chromatin mostly free or is it essentially covered in DNA binding proteins (even if non-specifically) at any point in time?
11
18
100
Sigh. The level of profound confusion in the scRNA-seq analysis field keeps surprising me. I've now run into people that, in a discussion about technical artifacts of scRNA-seq, where I claimed there is no evidence for zero inflation, first insist that it does exist, and then 1/n
@NimwegenLab
@maelouisewoods
@GorinGennady
@arjunrajlab
The existence of ZI as a "technical artifact" is distinct from divergence from a NB or Poisson dist. This paper e.g. argues that zero inflation is observed for certain genes & is of Biological origin (& recommends still using a GLM with NB counts w/o ZI).
2
2
15
3
13
99
Let's reflect on the fact that the current climate in the scientific community is such that whoever published this quite reasonable set of points felt compelled to hide their name.
5
23
97
Does anybody else find this hard to believe also? All those patterns with evolutionary rate and dN/dS rates as function of gene expression and essentiality were all mistaken? I really have a hard time wrapping my head around this at first reading.
Synonymous mutations in representative yeast genes are mostly strongly non-neutral - if the findings hold true more broadly we may need to re-visit many preexisting assumptions about mutations, selection, effective population size etc
9
64
216
13
9
95
Are you a microbe considering bet hedging but are worried about the costs? Amazingly, this simple trick can give you all the benefits of bet hedging without any of the costs.
Great work of
@daandegroot91
together with
@TjalmaAge
and Frank Bruggeman. 1/n
3
20
96
Do you have transcription factor ChIP-seq data? Give our completely
automated Crunch pipeline a try at:
and read about it here:
1/n
4
35
88
with its parameters, until one gets a picture that looks nice and agrees with prior biological knowledge and expectations of the authors. And then pretend in the paper that this picture somehow unambiguously represents the 'structure in the data'. In scRNA-seq talks one 7/n
2
1
79
looked very different just by tweaking some parameters (and I'm not even mentioning the preprocessing of the raw data!). But somehow we all seem to feel that because 'everybody does it' it is impolite to point out how meaningless these pictures are. 9/n
2
1
77
I don't know why
@mbeisen
deleted his account but I totally agree that the didn't say anything that requires him to be canceled and I find the calls for him to be fired and for Elife editors to quit 'en masse' distasteful. That said, I did find his tweet quite offensive. 1/n
I will probably end up regretting saying anything about this, but I do feel that it would be cowardly of me not to speak up on the cancellation of
@mbeisen
for retweeting the Onion article and expressing his opinion about the bombings and impending invasion of Gaza. 1/9
50
106
1K
4
4
75
I'm sure that in 20 years, the fabrication of important-sounding BS stories out of a carefully cherry-picked set of `facts' will still be a viable career option.
Yuval Noah Harari: we have no idea what to teach young people that will still be relevant in 20 years
1K
866
4K
1
2
73
Face it people.. Something as grandiose as 'laying out the landscape' just cannot be done succinctly.
Pity that this 'landscape' mostly resembles a giant landfill. There might be some perfectly fine results in there. But nobody wants to go into the stinky mess to look for them.
To the colleagues and friends who find our paper overwhelming, and instead prefer to write and read more manageable papers with more focused messages:
I respect your perspective and your approach to science greatly, and find it invaluable, of course. Both broad and focused
49
96
975
5
3
75
My institute is hiring! Broad search. We're also interested in quantitative/theoretical people. The Biozentrum is an incredible place to work. Great support, facilities, funding... and quality of life here in Basel. Please consider applying and RT.
0
48
73
If you're a scientist and not completely indifferent to the origins of this pandemic that killed over 10 million people, you owe it to yourself to at least once carefully read through the time line put together here.
The timeline on early behind-the-scenes discussions of the lab leak theory has been thoroughly updated.
17
132
414
2
17
70
Although it is certainly bad that contributions from Fouchier, Koopmans, Drosten were not acknowledged on the proximal origins paper, most relevant is that we now know their arguments were invalid (e.g. knowing the DEFUSE proposal etc.) Still waiting for a credible investigation.
David Fishman: “[It] would have made my spidey senses tingle had Fouchier been a listed author, knowing [his] apparent inability to consider the ethical dimensions of risky research...[It fits] a broader pattern of non-transparency and even subterfuge.”
5
17
83
1
20
69
Very informative thread. Showing once again that results of scRNA-seq analysis are highly sensitive to even small details of the procedures, in contrast to what seems to be implicitly assumed in many studies.
The choice of whether to use Seurat or Scanpy for single-cell RNA-seq analysis typically comes down to a preference of R vs. Python. But do they produce the same results? In w/
@Josephmrich
et al. we take a close look. The results are 👀 1/🧵
16
414
1K
1
11
70
For those of you interested, the proper Bayesian analysis is really simple: f =prevalence, s = specificity, r = sensitivity, p positives and n negative in sample. Info on test: fp false positives, fn false negatives, tp true positives, fp false positives.
Loud sobbing reported from under reverend Bayes' grave stone.
Seriously, I might use this as an example in my class to show how NOT to do statistics. Note that the CI on specificity includes false positive rates larger than the observed fraction of positives.
12
86
241
3
20
69
Covid19 killed close to 20 million people.
Avoiding a repeat of such a calamity requires that we find out what happened.
Given this, any scientist owes it to themselves to be at least aware of the facts laid out in this timeline.
5
24
69
Our >5 year collaboration with
@florianjug
,
@TheGeneMyers
and
@osilander
is finally out! Check it out, quantitatively track single-cell gene regulatory responses to changing environments. Including microfluidic chip design and fantastic tracking software.
2
47
67
or intelligible objective. Consequently, we have essentially no idea what 'structure' it captures and how this depends on its parameters.
The way it appears to be typically used in the scRNA-seq field is to run it repeatedly (Oh yes.. it's also stochastic!) and fiddle 6/n
2
2
65
@jbloom_lab
looked at the Chinese CDC data from the Huanan market some more and... no.. there's really no hint of evidence for animals infected with SARSCoV2.
I have performed additional new analyses comparing SARSCoV2 to other animal CoVs in environmental samples from Huanan Market
Paper w these new analyses is & full computational pipeline at
The new results are summarized below.
5
73
201
2
20
64
Do we need any more evidence that some virologists behave insanely irresponsibly and oversight of dangerous research is completely lacking?
Scientists at Boston University have posted a preprint of work they've done on a chimeric SARS-2 virus. NIAID, which provided some of the grant money, has questions about why it wasn't informed about work that might be viewed as gain of function science.
58
173
388
3
19
64
Zoonotic events are typically associated with an early burst of adaptation to the new host, which was not observed for SARS-CoV2. This is more consistent with a lab leak of a strain that has been passaged on human cells or in ACE2 mice. Then there is the furin cleavage site. 11/n
2
8
64
@jbkinney
And there's the rub. I believe many scientists realize that scandalous misdirection is coming from some quarters regarding the origin of the most deadly pandemic in a 100 years. But most think being taken for a right-winger is worse than staying silent about this scandal.
7
15
59
I recently also searched online for an official position of the Chinese government on this and couldn't find one. It seems the strategy is to frustrate investigations of possible origin inside China, play up possibility of import from abroad, but take no official position.
In recent coverage of pandemic origin, LA Times avers that "China insists that SARS-CoV-2 arose from a spillover event."
No link or reference. Got me thinking, what does 'China' (i.e. central gov) actually put forward as likely origin? Unclear 2 me.
14
4
27
8
12
62
Indeed. And I see little alternative than to conclude these scientists have been purposely trying to steer away the conversation from a serious investigation into a possible lab leak. Even now. After 20 million deaths.
4
14
62
Fully agree with
@jbkinney
. If you haven't bothered to follow the SARS-CoV-2 origins story, please take a look. I have been astounded at how little outrage there has been in the scientific community about the clear intentional obfuscation from top officials and some virologists.
This is a very compelling interview. Highly recommended for scientists who still dismiss the possiblility that SARS-CoV-2 might have been experimentally manipulated before accidental release. Also recommended is the
@PNASNews
article by Harrison & Sachs (who is interviewed here).
7
23
86
2
18
59
Well worth hearing for anybody that believes that "the scientific evidence" shows that SARS-CoV2 cannot have come from a lab. And for those of us in the scientific community, I think it should give us pause to note the number of occasions where 1/n
4
11
59
Sorry for the delay in replying. I ran into this in my own (unpublished) work > 10 years ago, assumed it was well-known, but I don't have a reference. However, I looked up my old notes and I think it can actually be derived relatively easily. So we will start from a set of 1/n
@NimwegenLab
This is very interesting. Do you have a reference you can share with an exposition of the result?
1
0
4
1
6
58
knows what UMAP actually DOES to the data. And that is no surprise either, because it is actually quite hard to find a coherent description of what the algorithm is. There's tons of pages on the internet that present folklore about how it 'preserves local structure' or that 2/n
1
3
59
Tell me you have never looked at how people analyze scRNA-seq without telling me.
Btw, for WGS it’s actually much more reasonable to just keep top PC comp. Cause there’s not the massive sampling noise that scRNA-seq has. And we know the first two PC components of individuals 1/n
Also... the All of Us paper did a PCA on the matrix of SNPs, took the first 16 PC of a high-dimensional dataset and then did UMAP on that? What?
8
5
72
2
6
57
No, it's really simple: the AF3 publication fragrantly violates the policies of Nature portfolio journals, thereby undermining them. There are ample avenues for 'publishing' whatever one wants. But there either are standards at Nature, or there are not.
This is a complex discussion that I am hoping will continue well beyond this letter. We need to find standards that work in an increasingly complex landscape of research and publishing, and that balances diverse interests.
We need to find solutions that enable science to ...
2
16
64
2
10
54
Unicorns! We have several positions open in my group, including two permanent research associate positions. The first is in our wet lab and will allow a researcher to pursue their own projects while contributing to other projects in the wet lab. The second is for a 1/n
1
30
55
@DavidBahry
@MichaelWorobey
Of course. But the whole discussion is surrealistic to me. It's obvious to any moron that the Chinese authorities have been strenuously frustrating attempts to trace the origins of SARS2. But somehow the real outrage is somebody noting and pointing out some small aspect of this?
4
12
53
But it took PhD student
@jeremiebreda
and myself several years to work out how to do it right. I'm really happy that we managed to get it out before the decennium ends.
2/n
1
7
53
@manoliskellis
I'm sorry but I honestly feel that this kind of shock-and-awe approach to science is doing terrible damage. I'm a theorist that has interests spanning multiple areas but the single-cell genomics/epigenomics field is by far the most dysfunctional that I have ever encountered. 1/n
1
6
51
Data analysis 101 lesson. What did you think when you saw that fit to the red data points? Turns out the report also provides the same data with the y-axis on a log-scale. What do you think now?
3
13
52
@_JanHartung_
Yes.. in what organism? Importantly, these things are highly context-dependent. We have developed tools to use transcriptomics or epigenomics data to answer those questions in a context-dependent manner:
1
6
51
I find is an astounding thing to say, given that everybody knows that investigation of a possible lab leak has been thoroughly obstructed. In contrast, possible zoonotic origins have been investigated for a year. And what evidence has that uncovered? Anything?
4
14
51
No! The future is data generators with such a deep understanding of analysis that they know how to design their experiments properly.
Decoupling of data generation and analysis is the future of academic research (mainly for better, sometimes for worse).
13
15
23
5
33
51
Gene expression noise and regulation are intimately entangled. Check out the new preprint by
@urchueguia_aran
.
I will briefly summarize some of the key things we find. 1. Using fluorescence flow cytometry with transcriptional reporters we measured 1/n
3
28
49
preprint is full of category theory talk that is impenetrable to most people, and that sounds like gibberish to people that actually know category theory (so I have been told). For those of you interested, I found this informative:
I'm not going to 4/n
2
0
48
Hmmm... If one were a suspicious person, one might see this as an attempt by the directors of NIH and NIAD to bring in two henchmen to pressure
@jbloom_lab
to alter or retract his preprint, and making sure to wash their hands of it when they notice the failure of this attempt.
I’d like to provide additional context to
@KatherineEban
’s article (), which reports on meeting w NIH prior to posting of my pre-print on Wuhan
#SARSCoV2
sequences deleted from NIH’s SRA. Meeting also in recent FOIAs: (1/n)
24
166
437
3
10
48
One thing I'm learning is that I have apparently been severely underestimating how hard it is to do data analysis. There's so much misleading, confusing and even nonsensical analysis of COVID19 data going around,
@richardneher
's below is like breathing clean fresh ocean breeze.
2
17
45
@Ayjchan
@MarionKoopmans
In Dutch he says 'we hear about such things early' and he mentions 'the days after sinterklaas, early december'. Sinterklaas is december 5. It's a remarkable boast for sure.
2
3
43
try and summarize but what is absolutely clear is that UMAP (like tSNE) performs a whole host of complex non-linear transformations to the data and that the objective function it optimizes is essentially impossible to interpret. It doesn't follow from any clear rationale 5/n
1
4
46
the title from `No evidence' to `Insufficient evidence'. I do not think this is a minor issue.
What is evidence? Since I've noticed that there seems to be quite some confusion as to what the term `evidence' means in scientific discourse, I'd like to point out there is a 2/n
1
1
45
As an
@snf_ch
grant recipient at a Swiss scientific institution, I'm not reassured at all by this bland response to the frankly disastrous proposals by their head of grant management. One can hardly think of an approach that would more effectively destroy Swiss science and 1/n
2
7
43
@boehninglab
I teach a ‘stats’ course and I utterly fail to understand how anybody can think one can teach stats to students that do not know derivatives/integration. Just for starters. How do you calculate a cumulative from a density? Or how do you get the maximum of a likelihood function?
6
0
44
make me believe that, whatever fitness effects are being measured here, they are mostly the effects of something ELSE than the point mutations in question. I'm pretty sure
@leonidkruglyak
is right about what's going on here. n/n
@TriggerLoop
@_mattrich
@RyanDhindsa
@RMeyerSchuman
having done these types of large-scale editing experiments, even the controls often look mildly deleterious due to various "off-target" effects, so without controls that capture every aspect of the manipulation except the introduced edit, there is no basis for any conclusions
5
16
88
0
4
44
Oh god.. this stuff makes me want to cry.
1
5
43
Apparently, just encouraging people to read a documented time line of events around the origins of COVID19 gets some people very VERY upset.
I don't know and never interacted with this person, but this is straight up slander.
10
4
44
claim that UMAP is better at preserving 'global structure' than tSNE (without ever defining what is meant by 'structure') but virtually none that give any concrete description. As far as I know, there isn't even an official publication on UMAP, only a preprint. But that 3/n
2
0
43
I like the new building
@biozentrum
. This is the view of the Rhine from my office right now.
2
0
43
@lpachter
Punchline is to teach young scientists, by example, that the purpose of a scientific poster is not to try and communicate any insights you might have gained, but to just 'shock and awe' whoever dares to come near it.
0
0
41
I have noticed this too and have come to believe that this really shows the lasting bad influence that R.A. Fisher's has had on reasoning in science. That is, this approach of just picking a null hypothesis that you find a priori most likely, and then sticking to it unless 1/n
However, I think that there is a philosophical divide among scientists in how to assess hypotheses that perhaps explains some of the gap in opinion. Ie, is zoonosis the "null" hypothesis that we need significant evidence to reject or are we comparing two competing hypotheses? 3/8
10
33
259
6
8
42
@martinmbauer
My favorite in this genre is that an object with the shape of a Cauchy/Lorentzian distribution has no center of mass.
2
1
43
They claim they can tell human population size one million years ago to within 10 people? Is this a joke?
A new technique analysing modern genetic data suggests that pre-humans survived in a group of only 1,280 individuals.
18
140
388
2
4
40
The only known sarbecovirus carrying one as far as I know. With human codon bias. Such sites have been engineered into viruses in the past and we know that the WIV proposed to insert such FCSs in novel coronaviruses. Thus, this feature too is more consistent with lab origin 12/n
3
4
41
the preprint. Now, one might argue all kinds of reasons for this, but I do think these numbers are telling. So what, exactly, does the publication in a reputable journal really add? Did the review process and revision fundamentally alter the work? No. 5/n
1
6
40
Addendum to this great thread. We recently showed that, upon switches from glucose to lactose, the majority of single cells experience a delay because their lac operon expression is at zero and, consequently, lactose does not get into the cell and/or is not isomerized into 1/n
Prepping for a class session on the Lac operon in Molecular Biology and it prompts my yearly reminder to you all that what we were taught about the actual substrate of the Lac operon is almost certainly wrong...1
22
276
1K
3
9
39
Cynic that I am, I wasn't so shocked by how common fraud is in scientific publications. But the fact that 2/3 of the time, journals just completely IGNORE iron-clad proof of major problems (if not outright fraud) after it is specifically pointed out to them. That shocked me. Why?
Our first keynote address is from Elisabeth Bik (
@MicrobiomDigest
) who is giving us an insight into her work scrutinizing integrity in scientific publications. 🔎
Elisabeth will also join students for an informal session with students tomorrow.
#bc2basel
1
8
28
5
10
38
of European descent give a rough map of Europe. Which also makes sense because the top PC are driven by high frequency SNPs.
It’s funny how when standard scRNA-seq analysis is applied in the genetic pop struc field, all of a sudden people realize it’s nuts. n/n
0
3
37
@manoliskellis
It is almost all make-belief IMHO. Hugely complex layers of unmotivated ad hoc transformations and filterings applied to the data.. ad hoc measures and statistics with murky meanings invented in virtually every paper. There are almost as many analysis procedures as there are 2/n
1
3
39