Scoop: new data-set of the world’s most-cited researchers finds hundreds who’ve amassed more than half their citations (!!) from themselves or their co-authors. By me and
@DalmeetS
.
Scientists working with the WHO have acknowledged errors in their COVID excess deaths study.
They revised excess death figures for Germany, down 37%, & Sweden, up 19%.
Germany now well below UK. Not yet on WHO official site: update later in year.🧵
Weird. Author of Nature paper
@PatrickTBrown31
says he looked solely at effects of warming on wildfires, ignoring vegetation and human ignition pattern changes, to mold research to journal's desire. Yet peer-review file shows issue raised in review & authors argued against it!
Last week, I described our paper on climate change and wildfires:
I am very proud of this research overall. But I want to talk about how molding research presentations for high-profile journals can reduce its usefulness & actually mislead the public.
Some researchers say that at least a quarter of randomized clinical trials in some fields are untrustworthy (perhaps fake?!), but still cited in reviews that guide medical practice.
How big is this problem & what's the evidence? Here's my look in
@Nature
How large is science's fake-paper problem?
Massive, finds a new ML text-analysis by
@ClearSkiesAdam
. It suggests paper mills rise to 1.5-2% of all papers; 3% in biomed.
That's 70,000 papers last year alone, >400,000 overall. As a 'conservative estimate'.
Fascinating to see how to use GPT-3 [ChatGPT's predecessor] to edit a scientific manuscript (adapted from
@miltondp
's January preprint, for this
@nature
article about generative AI and science )
Below: what Lancet editor Richard Horton wrote on 24 Jan, the same day a report from China was published in the Lancet. Now Horton lambasts scientists and government for not taking the warnings of this Lancet report seriously enough. Yes ... but at the time, he urged caution too
A call for caution please. Media are escalating anxiety by talking of a “killer virus” + “growing fears”. In truth, from what we currently know, 2019-nCoV has moderate transmissibility and relatively low pathogenicity. There is no reason to foster panic with exaggerated language.
@jburnmurdoch
is it possible you can separate out Covid's lethality to unvaccinated people vs to vaccinated people, as comp'd to flu? Cos this averaging may lead unvaccinated people to think 'it's no more lethal than flu'. Which, for them, is wrong.
On the pandemic's true death toll. Certainly millions more than official counts, but how many millions, and how is this work done? by
@davidneiladam
(1/4)
In 2018, two scientists let Nature reporters chronicle their lives as they launched their first labs. We all had no idea of the global and personal crises to come. Here's their story, with many thanks to
@danbose
and
@dozenoaks
for their generosity.
"Influenza, polio and more have shown that infections can change lives even decades later. Why the complacency over possible long-term effects of COVID-19?" by
@lfspinney
in
@nature
A big development for text-mining?
@carlmalamud
has released online a giant, free index of 355 billion words and sentences in 107 million science papers. By
@hollyelse
in
@nature
Delighted to win 'Editor of the Year' award from the Medical Journalists' Association last night -- many congrats to all winners & shortlisted. And amazing to be at a real-life celebration of journalism again (negative-flow tested, indoor-masked)
@mjauk
Editor of the Year (a fiercely contested award) - winner
@richvn
features editor
@Nature
for The race for coronavirus vaccines: a graphical guide, The simulations driving the world’s response to COVID-19, and A guide to R - the pandemic’s misunderstood metric
#MJAAwards
2021
Survey in
@nature
: of >9,000 scientists who are or have been on Twitter, more than half said they've cut back their use in the past 6 months and 7% said they've stopped using it. More details in story by
@myriam_vidalv
There's too much booster boosterism right now. As
@ewencallaway
explains: "Scientists say that the case for COVID-19 vaccine boosters at this point is weak. They might not be necessary for most people, and could divert much-needed doses away from others".
A intriguing preprint using AI to try to measure the 'thoroughness' and 'helpfulness' of 10,000 peer-review reports came out in July: I talked to author
@annasvrn
about it + other efforts to study peer-review at vast scale. |
@nature
Another profile of Elisabeth Bik,
@MicrobiomDigest
, this time in the New Yorker (if you want to try how hard it is to spot such images yourself,
@nature
's profile from last year has a quiz:
A new study (this one peer-reviewed, fwiw) reaffirms what earlier work has found: excess deaths during pandemic 3x higher than official COVID deaths. By
@davidneiladam
Best thing about this joke-paper-sting-to-expose-crappy-journal is the peer-reviews, which were made available after the paper got retracted. Editor who didn't want the paper published was replaced by others who were ok with it.
Researchers have analysed anonymized phone records of tens of millions of people in low-income countries under the rubric of “data for good”. But do the benefits outweigh the risks?
@amymaxmen
investigates
@Ramblingproses
@carolecadwalla
@BBCNews
The WHO's Mike Ryan has condemned it (e.g. see below for one of many media articles covering Ryan's comments and the letter). Agree that the BBC has been a government-stenographer on it.
Scientists have identified a host of genetic variants that are linked to an increased risk of developing severe COVID-19. The variants affect processes ranging from immune-system signalling to blood clotting. by
@FKreier
@TomChivers
Two chasers to this:
1) Even if you do really have antibodies, you may not have long-lasting immunity to COVID-19. (We simply don't know).
2) UK analyses say kits are nowhere near '90%'; more like 50-60%, for mild cases. ( )
Here I explain the big changes announced yesterday by the open-access
#Plan_S
. Lifts restrictions on where Plan S scientists can publish -- as ensures that they can share accepted manuscripts immediately online, CC-BY, even if a journal doesn't want this.
Researchers at Meta (formerly Facebook) have used AI to predict the structures of some 600 million proteins from bacteria, viruses and other microbes that haven’t been characterized or (mostly) cultured in labs. by
@ewencallaway
|
@Nature
Ugh in lots of ways. First, there are papers describing using traditional chinese medicine against COVID-19. And now it turns out one paper seems to have copied its CT-scan images from another. Sounds like the tip of a very murky iceberg.
Wow.
The paper on healing a COVID-19 patient with Traditional Chinese Medicine that I critiqued on PubPeer appears to have copied CT scan images from another paper by another group.
See:
and
Scientists studying highly-mutated B.1.1.529 coronavirus variant spreading in S Africa, likely termed 'Nu' variant soon. Qs: can it evade vaccines, does it cause more or less severe disease, and does it spread faster than other variants |
@ewencallaway
@DoomlordVek
@ProfSunnySingh
Tbf the series is many things, IMO. It's all at once cliched, formulaic, copies predecessors [both 'school stories' and 'magic'], poorly written, ridden with imperial/classist/racist overtones --- and yet also a rollicking page-turner.
Heartily recommend this in-depth profile of Elisabeth Bik, science sleuth extraordinaire
@MicrobiomDigest
, by
@HelenShenWrites
. With a few tests inside the story: can you spot the image duplications that Bik sees?
Lancet study retracted - a reminder (if we needed it) that a 'peer-reviewed' paper can be as flawed as a preprint. Still unclear what happened during the Lancet peer review process here. It'll be interesting to see if NEJM, which also published a Surgisphere paper, follows suit.
Here's a nice feature explaining how what scientists have learnt about glycans - the sugar-based polymers that coat cells and decorate most proteins - fundamentally hinges on Carolyn Bertozzi's bio-orthogonal chemistry.
#NobelPrize2022
An IEEE journal editor saying they don’t need to check out allegations from non-institutional or anonymous tipsters. I thought this attitude had died 10 years ago! (It’s certainly against COPE guidelines).
The
@IEEEorg
Editor in Chief wrote me back:
"I am not interested who you are. I do not need to tell you what is justified. We do not deal with anonymous allegations or allegations coming from non-institutional email addresses"
(I write from my gmail address; I am not employed).
@alexwitze
on James Webb Space Telescope's home for the next 20 or more years --- orbiting around the second Lagrange point.
'Webb telescope reaches its final destination far from Earth' |
@nature
Delighted to get an
#AAASKavli2021
science journalism silver award for my feature on ethics in facial recognition research () and congratulations to all the winners!
Grim and incredible story. US high-school kids churning out 'research papers' [enabled by pay-to-publish journals] to try to get into US colleges; papers can be of dubious quality.
I've reported for years on the influence of wealth on college admissions. This time I found a new twist: a bevy of online services that, for hefty fees, help high schoolers conduct and publish research to impress colleges.
Omicron blindspots: why it’s hard to track coronavirus variants -- by
@amymaxmen
|
@nature
. Includes this graphic showing delays between when a sample was collected and when sequence made public at GISAID
Re Facebook's new name (Meta).
Interesting to remember that a startup called Meta, which aimed to revolutionize scientific discovery with AI, was bought by the Chan Zuckerberg Initiative in 2017, then slowly mothballed.
Guess it's a popular name.
Chemists have invented ways to insert, delete or swap individual atoms in the cores of molecules. “Almost magical that these changes are now possible”, Richmond Sarpong (
@SarpongGroup
) tells
@markpeplow
|
@Nature
🧵on skeletal editing & why it's exciting
Plus, it isn't just the SAGE committee that's secret. Also secret are full minutes and membership of scientific sub-committees advising SAGE, called SPI-M and SPI-B. (Weirdly, the third sub-committee, NERVTAG, does have its full minutes and membership online. Go figure!)
I had no idea about this until yesterday. Unbelievable that the government claims to be 'following the science' but won't even say who the scientists are..
Google Scholar's 'Public Access' tab now tracks whether your papers should be free to read (because a funder requires it) and how many of them actually are. I saw some criticism, so asked GS co-founder Anurag Acharya to explain how it works.
After his retirement a scientist (whom we reveal here as Kuo-Chen Chou) racked up 50,000+ citations to become one of world’s most highly-cited. He’s just been barred from an editorial board for citation manipulation. Investigations ongoing elsewhere.
AI -> scientific search: two big science databases, Scopus (Elsevier) & Dimensions (Digital Science), announced ChatGPT-based LLMs. Web of Science (Clarivate) plan similar. The concept is not like Meta's Galactica though.
My report for
@Nature
:
"In the past half-year, AlphaFold mania has gripped the life sciences."
@ewencallaway
on how software that can predict the 3D shape of proteins is already changing biology.
"Fusion research has shifted from gargantuan state- or internationally funded enterprises to sleek, image-conscious affairs driven by private companies, often with state support".
A tour of the nuclear fusion hopefuls, by
@philipcball
|
@nature
"Russia’s unprovoked invasion of Ukraine has unleashed an outpouring of condemnation from scientists and research organizations worldwide" incl collabn' freezes; calls for boycotts of Russian science. by
@HollyElse
New story: I've learnt of at least four journal publishers adopting AI software to automatedly flag duplicated images in papers, pre-publication. Still needs editorial oversight, but an early taste of how image-screening AI cld sweep through sector. (1/3)
Future quantum computers (if they reach full scale) will be able to break Internet security; this year NIST will recommend stronger encryption algorithms. Then comes updating security protocols; tests haven't all run smoothly. By
@dcastelvecchi
|
@Nature
VC and other private investors chucked around half a billion dollars at quantum tech firms in 2017-2018, finds
@LizzieGibney
in this breakdown of who’s funding what, where, and why.
More than 50 publishers, representing 15,000+ journals, are preparing to ask authors, reviewers & editors globally about their race or ethnicity.
@HollyElse
and
@j_perkel
dig into the story of this initiative to chart researcher diversity |
@nature
Should add that while these fake papers are [hopefully] ignored by most scientists, they do in fact get citations -- because they're citing each other -- and some do get aggregated into systematic reviews, particularly in the medical literature.
How large is science's fake-paper problem?
Massive, finds a new ML text-analysis by
@ClearSkiesAdam
. It suggests paper mills rise to 1.5-2% of all papers; 3% in biomed.
That's 70,000 papers last year alone, >400,000 overall. As a 'conservative estimate'.
The NASA probe crossed into the Sun’s atmosphere at 9:33 a.m. UTC on 28 April ... several months for scientists to download and analyse the data to be sure that the spacecraft had crossed the boundary, known as the Alfvén surface. from
@alexwitze
Researchers are building a human ‘pangenome’ that would represent the entirety of human genetic variation. But not everyone is ready to sign on. by
@rkhamsi
|
@Nature
“Scientific misconduct of the highest order” —-> or as one reader has emailed me: “my jaw is still on the floor as to how he was allowed to get away with such things for so long”.
"The World Health Organization (WHO) has quietly shelved the second phase of its much-anticipated scientific investigation into the origins of the COVID-19 pandemic" | by
@SmritiMallapaty
,
@nature
Le sigh. Paper shows that GPT-4 w/ Python can be used to “fabricate data sets specifically designed to quickly produce false scientific evidence.”
As
@cragcrest
explains at
@TheLWON
. Unsurprising, but still concerning.
(paper: )
Five EU countries have proposed sweeping restrictions on PFAS 'forever chemicals'. These 12,000+ chemicals are used in microchips, jet engines, batteries, air con and much more. Could the world find a way to go PFAS-free?
XiaoZhi Lim reports in
@Nature
What happened?
First, a reminder: the WHO estimated excess deaths, meaning mortality above expected levels, in the first two years of the pandemic.
Their global estimate (13.3 – 16.6mn excess deaths) was a little more conservative than other studies.
These data are not generally available. But there are some studies. Here, for example (from
@jschoeley
et al) UK has lost more life expectancy (calculated using age-specific mortality rates) than Spain & Italy, tho similar on crude excess death rates.
"Research-integrity sleuths have uncovered hundreds of online advertisements that offer the chance to buy authorship on research papers to be published in reputable journals."
@HollyElse
on the work led by
@nickwizzo
and
@AbalkinaAnna
First authors of this study are from Samsung AI Center in Moscow (launched 2018). Interesting to see with current scrutiny on which int'l Russian science partnerships still operating after Ukraine invasion. (Samsung suspended shipping products to Russia in March.)
Am very excited to share our new work on realistic avatars, which was recently accepted to ACM Multimedia! 😊 Huge props and congratulations to Nikita
@NikDrob23
, you did an amazing job! More results on our project page:
Excellent critique of a non-verifiable study on whether more tweeting leads to more citations - the authors won't share their data and say it's impossible for anyone else to reproduce their exact analysis.
New evidence from Australia on how a disturbingly large % of medical researchers fail to disclose pharma funding on papers (similar to findings in US).
News tidbit: retrospective testing has found Omicron in samples collected 1-3 Nov from individuals in England, Nigeria, USA: *earlier* than first known Sth. Africa samples.🧵
It's noted in this story on Omicron's origins, by
@SmritiMallapaty
|
@nature
After 8 years in peer-review, Mochizuki’s 600-page proof of the abc conjecture, one of the biggest open problems in number theory, has been accepted for publication.
@neurograce
@PatrickTBrown31
fwiw, there's also this in the supplementary file: the authors say they believe their design to hold other potential predictor variables (besides temperature) constant 'results in conservative estimates' of change in risk.
Wellcome Trust's updated open-access policy: from 2020, no embargoes, and won't pay for hybrid journal APCs [but won't outright bar publishing in hybrid journals]. Wellcome & the Gates have now joined Plan S -
But Germany's excess death figures looked too high. Why?
@PaulMainwood
,
@Martin__Sauter
and others noted that WHO’s *expected* deaths for Germany seemed too low. This seemed to be because scientists used a ‘spline’ extrapolation, too sensitive to a 2019 dip in German mortality.
@DanJCondon
@Nature
Excellent media scepticism & great question! Interesting though how the replies [including your own,
@DanJCondon
] focus on the research papers in Nature ... a topic irrelevant to the magazine content being shown.
Factcheck: Actually, most US PhDs still go on to academia (53% in 2022; 31% to business).
The charts shared were for data that excludes all the PhDs that go onto postdoctoral positions -- i.e. a hefty chunk of the academia! (I checked with NSF).
I analyzed 20+ years of NSF survey data.
Since 2015, academia has been the "alternative" career path for PhDs.
But it was the breakdown by field that really shocked me.
Do these graphs match your PhD experience? 1/
Incredible and worrying story. I’m left wondering: if orig. studies are proven wrong/fraud, wld this wipe out only Aβ-56 oligomer theory, or the ‘toxic oligomer’ thesis in general, or the Aβ thesis in general? (Given later research exists). Any Alzheimer’s experts have a view?
Were two separate, major lines of Alzheimer’s research tainted by image fabrication, with far-reaching implications for the field? I take a deep look for
@ScienceMagazine
🧵 1/11
well that's one way of looking at it!
"...ChatGPT’s currently free offerings could help put [essay mills] out of business. At the very least, low-income students will have access to the same opportunities for academic misconduct as well-to-do students have had for years."
Same issues affected Sweden (as
@COVID19actuary
has noted): adjusted raw data, and the spline extrapolation.
Reverting to original data, linear extrapolation, raised excess deaths for Sweden by 19%.
"Doudna was “really sound asleep” when her buzzing phone woke her and she took a call from a Nature reporter, who broke the news." what a moment
@heidiledford
It’s
#CRISPR
. Emmanuelle Charpentier and Jennifer Doudna ,who pioneered the revolutionary gene-editing technology, are the winners of this year’s
#NobelPrize
in Chemistry.
#NobelPrize2020
More on how scammers pretended to be scientists to get 'guest issues' stuffed with sham papers. Hundreds of retractions and more to come. by
@hollyelse
|
@nature
Wow, the concern over 'special issues' fraud at Hindawi is really hitting Wiley's financials. (Have research publishers seen their bottom line hit by fraud concerns to this extent before?) The firm's share price is currently down 18% after this update.
Scientific sleuths like
@gcabanac
are picking up dishonest use of ChatGPT in peer-reviewed papers.
Authors didn't disclose they'd used ChatGPT, but didn't delete the phrases 'As an AI language model...' or 'Regenerate response'. 🤦
By
@gvconroy
|
@Nature
Second, of course, the estimates all come with uncertainty ranges (or 'credible intervals' - Bayesian-style statistics used here). Here's one way to look at that for some countries' excess death rates (revised figures are shown here). Thanks to Victoria Knutson for this graphic.
An effort to sum up diversity [race/ethnicity] in UK science in one chart. That drop in representation for Black science academics is stark. From by
@LizzieGibney
|
@Nature
Signs that politicization of US–Chinese science, & the pandemic, is affecting research collaboration?
In 2018, there were more than 15,000 authors on research papers who declared dual affiliations in the US and China. That fell to below 12,500 in 2021.
@SmutClyde
talks to
@HollyElse
about how he searches for fake papers - including detecting possible paper mills by finding recycled networks of bogus references - & why he uses that pseudonym.
E.g.: on revised figs, Germany’s deaths in the pandemic were 6.5% above normal, while in Sweden, they were 7.5% higher! Even tho Germany's per-capita excess death rate is higher...
"predatory journals are re-issuing — seemingly on their own initiative without any consent — actual, peer-reviewed articles that have been published elsewhere."
Researchers reverted to raw data, and a linear extrapolation. Result: 37% lower excess deaths for Germany: from c.195,000 to c.122,000. (Ironically, using a spline extrapolation on raw data also worked ok, says
@jonwakeblue
).
If u're amazed by 'most-cited' researcher Rafael Luque churning out 58 articles thus far in 2023, check out other Clarivate highly-cited scholars, eg
Ben Zhong Tang: 97 articles (>1 every 24 hrs)
Pau-Loke Show: 90 articles
Kamlesh Khunti: 76 articles
(GS counts)
One of world's 'most-cited' scientists, chemist Rafael Luque, sanctioned by his institution, U Cordoba, for his ties with Saudi Arabia, Russian institutions. Reporting by El Pais.
Much speculation on whether Houghton will accept the Nobel, as he declined the Gairdner prize in 2013 for hep C because - rule-of-3 restrictions - it didn't include colleagues Quo and Choo. (But he accepted the Lasker in 2000 after 'agoniz[ing]' about it)
There may be a fascinating controversy in store for the Nobel Prize this year.
One of the awardees, Michael Houghton, previously was awarded but declined to accept the Gairdner Prize because it didn't include two researchers who he felt made key contributions to the discovery:
@garrett_wollman
@audrawilliams
@fanf
Surely the answer is to read better journalistic non-fiction, not to stop reading. Unfortunately it's very hard to know who's better - book reviews almost never distinguish between the Gladwells (👎) and the Quammens (👌) of this world. You have to find your favourites.
Counterpoint:
@noisemyulchi
points to a Korean interview where Fields Medal winner June Huh says that his wife, a fellow mathematician, gave up study when their child was born owing to his (laughs) 'insincere participation in childcare' (GTranslate).
The guy who just won the Fields Medal – like a Nobel Prize for mathematics – has a work ethic we should all emulate. Three house of focused work per day based on following his desires at that moment.
What time is it on the Moon? There is no standardised official lunar time - but we’ll need one when multiple lunar missions work together.
@LizzieGibney
|
@nature
reports how scientists are starting to work out a system.
This flowchart is probably too long to be worth tweeting, but it captures the key factors that are likely to lead to the coronavirus becoming endemic. From