Friends tell me I take
#COVID19
too seriously, it doesn't kill people our age, ask why I'm social distancing, etc.
I do this all to flatten the curve. I do this as a service to older folks and those with compromised immune systems. I do it because I think it's our moral duty.
@LTF_01
Sorry to hear. I’ve had very similar experiences with neurologists. One neurologist was so keen to reach a particular diagnosis he just kept asking me the same question again and again.
I am not a fan of this meme format mocking people for having read popsci books. It’s elitist in-group signaling.
It’s quite commendable when someone’s curiosity leads them to explore a topic in their free time. Many scientists, including me, were inspired by popsci books.
The longer I'm in academia, the more I see a resemblance between our field and professional sports. But I think folks in pro sports understand and embrace the competitive aspect and difficult parts of training in a more honest and healthier way. This advice resonated:
It's pretty awesome to think in the future, there will be coevolution between weeds evading detection by mimicking crops, and machine learning algorithms improving to detect them.
I've been thinking about Andrew Wyeth's 1948 painting "Christina's World" and how art can enrich how we teach genetics. At first glance, it features a woman sitting in a pleasant field. Yet look closer, and you notice tension in her arms and hands.
US universities ranked by number of "best scientists" (h index>155). Some under- and some overrated unis?
Harvard 69
MIT 21
Johns Hopkins 20
Stanford 20
UCSD 19
Caltech 13
Berkeley 11
Columbia 10
Yale 10
Cornell 8
Penn 8
Princeton 5
U. Michigan 5
U. Wisconsin 5
U. Illinois 2
Something I wish I knew at start of my PhD: sitting destroys your body. Unless you're actively doing strength training and mobility exercises to combat this, you will suffer aches and pains due to shortened hip flexors, weak glutes and hamstrings which can lead to backpain, etc.
I feel like as a postdoc, the hardest thing about science is never the science itself, it’s the existential dread and self-doubt about finding an academic job.
So
@AdamRutherford
is receiving a lot of skepticism about his tweet saying that if Cheddar Man has any genealogical ancestors today, he is the common ancestor of all present-day humans. Adam is correct about this — I think this is such a cool finding, so I'll explain it below.
Ok look, there just isn't any good evidence of transgenerational epigenetic inheritance being caused by trauma in humans.
I'll wager $200 — show me good causal evidence (we can pre-negotiate terms) and I'll pay you if you prove me wrong.
So why is this making it into Science?
In the 2000s, investigators found a pattern among a lot of scattered crimes: the same woman's DNA kept turning up at the scene. There were about 40 crimes total, including 6 murders, and the suspected serial killer became known as "Woman Without a Face" or "Phantom of Heilbronn".
Just incredible. The role significance testing has in the public (mis)understanding of science is immense. Any claim is seen de facto true if found to be statistically significant. Maybe we should just increase the (uncorrected) threshold to far below 0.05.
These crimes were all across Austria, France, and Germany. In 2009, investigators found her: she worked at the factory where the cotton swabs used for collecting DNA were made. Moral of the story: never discount contamination as the cause.
A large fraction of faculty act as editors and peer reviewers for journals. Setting the precedent that their public personal views or comments somehow represent the journal is in my view a very bad move. It will effectively open up a new avenue to silence academics’ speech.
Alright, since folks are commenting on the Guardian article calling Bayes Theorem "obscure", here's a story about doctors, Bayes Theorem, log scales, and learning and listening from people even when they don't know the thing you know.
Cultural Revolution in China, Iranian Revolution, Hitler Youth in Germany, Khmer Rouge (which led to the Cambodian genocide) all had student support against the ruling class.
University activists must actually study history at their universities, and not just engage in activism.
A good law of history is that if you ever find yourself opposing a student movement while siding with the ruling class, you are wrong. Every single time. In every era. No matter the issue.
It's remarkable how evil the two body problem is: it forces a couple to choose between living apart to pursue the best opportunities so *hopefully* they can get the chance to live together in the future, or sacrificing those opportunities to be with each other in the present.
This Lexis plot depicting UK mortality across the 20th century is wild. I had no idea cohort effects (diagonal 'scars') were so strong. Events like the 1918 flu led to not only high annual mortality, but an increase in mortality across one's lifetime. HT:
@salonium
's newsletter.
This is a good move by Stanford in my opinion. It suggests the recent trend of universities commenting on news events was a mistake, and it will issue fewer positional statements.
I think the sad actual answer to "What is the biggest open challenge in biology?" is:
1. Getting people to share data.
2. Structuring, organizing, and annotating data with metadata so it's useful.
3. Building higher-level abstractions so people can efficiently work with big data.
I'm teaching my partner bash. The biggest "wows!" have been:
1. control-r, type part of a past command, control-r again and again to cycle through history.
2. cd -
3. tmux for work on a server.
4. control-a / control-e / option arrow for jumping around.
5. hitting tab a lot.
I am once again asking that biologists recognize that human and legal rights do not come from sex chromosomes systems, anisogamy, or parthenogenesis in whiptail lizards. Do not get pulled into debates that conflate the two in a motte-and-bailey setup.
Holy shit: "A bug in the seaborn 0.11.2 plotting software [3], used by Park et al. [1], silently drops the largest data points in the histograms."
Yikes. This is why we need more replication.
Wow. A bug in the Seaborn data visualization software hid many CD=1 papers, leading Park et al to incorrectly conclude that disruption in science and technology is declining (top histogram), while it is not ( bottom histogram).
@VincentGinis
Being a postdoc is working pretty much non-stop on both actual science, and suppressing the feeling that it will be all for naught if you never make it into a permanent job.
Shall we dissect how completely absurd the claims in this tweet are? It's a lesson in how scientific work can provide a thin veneer of plausibility, which is then packaged with more absurd claims in a classic motte-and-bailey. This tweet has half a million views!
So much of bioinformatics is endless writing of parsers for file formats written by software, and it just shouldn't be this way. This is needlessly time-intensive, bug-prone, and prevents higher-level abstractions from being developed.
We need tidy bioinformatics.
I keep seeing this take and it's annoying me: defending RA Fisher's eugenic beliefs because of his era is *not a good historical argument*. Wright stayed clear of eugenics despite it being pervasive among all leading geneticists at the time (Davenport, East, Castle, Fisher, etc).
I'm very pleased to announce that I'm starting a postdoc with
@ras_nielsen
at UC Berkeley! I'm headed back to California and across the Bay from where I grew up.
The hardest part about academia is postdoc paper production rates required to get a job mean every little obstacle encountered —a normal part of doing science— feels like a death sentence.
I've used this tidyverse data analysis pattern for reading in multiple files' worth of data and extracting metadata four times in the last week, and taught it to two people. Decided to document it here:
Something I noticed about the bogus Shi Huang paper on genetic diversity and cognition: even though it's a real paper, it's so bad that it sort of functions as a Sokal hoax for race and intelligence researchers. They fell for it, which tells you a lot about their "science".
A mutation cannot "see" its fitness effect — this is a central tenet of evolution that has not been refuted. This is saying there is causal independence between a mutation and its fitness effect. The causal arrow only points one way.
Here's a quick introduction to Snakemake I wrote (including a short section on good old fashioned Make). It tries to clarify some aspects of Snakemake I found confusing early on, and builds up to a SLiM example.
My new preprint, "Why do species get a thin slice of π? Revisiting Lewontin’s Paradox of Variation" is up on BioRxiv! Now that I’ve had ☕️, I’ll step through some of the major points in this paper. Link:
It's mind-boggling to me how complex bioinformatics has gotten. I think modern software that abandoned the long-established tradition of Unix-centered CLI design was a giant mistake, and we're paying for it.
Something I'm reminded of as I'm tromping around in the bioinformatics weeds: every genomic estimate of every parameter is conditioned on a ton of arbitrary filtering settings, which for the most part, we choose based on a combination of precedence and vibes.
Supplementary materials should exude warmth to any reader who dares venture through them. "Dearest reader, friend, explorer, I'm sorry some complex set of events in your life have brought you to this section..."
I think what Musk doesn't understand is that free speech in online spaces isn't hard. 4chan has free speech, but scientists, musicians, politicians, etc don't want to hang out there. Community building in online spaces is much harder and involves incentivizing good behavior.
I'm pleased to share a beta version of SciDataFlow, a command-line tool to track changes to data in research projects, push and pull data to remote repositories like Zenodo and FigShare, concurrently download tons of data, and more!
Another day, another conclusion based entirely on a horribly biased sample in the medical literature. Sampling marathon runners to show there’s no risk of knee issues associated with running is like sampling 90 year old smokers to investigate the impact of smoking on longevity.
This strikes me as a very bad move by
@eLife
. It's wrong to terminate someone for voicing their political views on their personal account, which clearly does not represent the views of the journal nor any other institution an individual is affiliated with.
I have been informed that I am being replaced as the Editor in Chief of
@eLife
for retweeting a
@TheOnion
piece that calls out indifference to the lives of Palestinian civilians.
In my book, I share what I call the golden rule of bioinformatics: "never ever trust your tools (or data)". This reddit thread on scikit-learn is yet another reminder why.
Overall, I think the view that science is somehow at odds with art, or poetry, etc. is narrow. Art can provide depth and context to an otherwise technical paper or subject in a way science alone cannot.
At what career stage can you just put weird shit in your acknowledgments to confuse future scholars, like "The author would like to thank Fleetwood Mac for their album 'Rumors', without which this study would literally not be possible."
In 1962, Dobzhansky wrote more clearly about human ethics and biology than most stuff I've read in the last few years. Thanks
@kph3k
for sharing this masterwork.
I find these types of arguments incredibly weak and incautious. Don't use evolutionary biology to back human rights or argue for certain laws. What non-human animals do has little to do with how we choose to shape our laws and society.
I bet some bioinformatics results have merge errors caused by bad joins on variant positions —
@jgschraiber
and I were just discussing this. Why? Because VCF is 1-based, BED is 0-based, GTF/GFF is 1-based, BAM is 0-based, SAM is 1-based. Joins on (chrom, position) key are bad.
Alphabetical order mismatch and 52 of 78 neighborhoods had wrongly merged data. I spend a lot of time teaching advanced inference methods, but boring research data management remains the most essential skill. And that includes auditing for merge mistakes.
I'm pleased to share an alpha version of GRanges, a Rust library for working with genomic ranges and their associated data.
I wondered if it was possible to write a library that:
1. Has tidyverse-like syntax for genomic data manipulation.
2. Is extremely performant.
It is.🚀
Most evidence that eating breakfast leads to weight loss was funded by cereal companies; in one case, Quaker even designed study and edited the manuscript! Meta-analysis (visual abstract below) supports idea that skipping breakfast leads to weight loss.
A study in the Journal of Environmental psychology found that being near water reduced blood pressure and heart rate. The participants? Volunteers from the Davis Aquatic Masters swim team. Talk about a non-representative sample 😅
I’m impressed academia collectively seems to be listening and learning, but is it just me, or do these circulating twitter pledges seem performative and generally a bit odd?
Random reflection this morning: it should humble every scientist that from 1992-2005, this was the official USDA nutrition guidelines for ~300M Americans. Supposedly based in science, victim to immense lobbying, and still many Americans abide by its utterly wrong recommendations.
My recent paper on Lewontin's Paradox is live on
@eLife
! I will write a blog post explaining it to a general audience soon, but you can check out the accepted version PDF here: Much thanks to Guy Sella,
@mwpen
, and two other reviewers for helpful feedback!
Re-running some code on a new machine caused one of my figure's points y values to inexplicably increase by a tiny bit. I mean shit, I wrote a book on reproducible bioinformatics and some days I just think it's all impossible and we shouldn't trust anything.
The funny thing about science is you spend thousands of hours working on something, and inevitably about 90% of the way in you think, "there are a dozen ways I could have done this better if I knew what I knew now".
The subject of this painting is Christina Olson, Wyeth's neighbor and friend. Olson developed a degenerative muscle condition and preferred crawling —as depicted— rather than a wheelchair. The MOMA has a great video exploring this piece and disability:
Can we please get a billionaire to finance making a native tidyverse in Python. I'd wager global productivity in data science would increase 2x and frustration down 30x.
There really is no excuse for why Python packaging is so atrociously awful. R does it better. Rust does it better. Hell, even Lua-based Vim extensions work better. How is it possible it is this bad? Millions of collective hours wasted fighting this.
Following the COVID-19 on
@nextstrain
has been a really fascinating way to see this evolve in realtime, and
@trvrb
deserves a huge amount of praise for keeping us updated. But everyone should also take note just what a stellar example of open science Nextstrain is 1/3.
@sapinker
@Harvard
Every single one of these figures has a y-axis that is "percentage of white respondents". All you've shown is it's become less accepted to be openly racist through time (and we know polls on this are not accurate: see the Bradley Effect! ). Come on...
My latest with
@Graham_Coop
, describing how polygenic selection creates autocovariance in neutral allele frequency trajectories. With temporal genomic data we can measure these covariances, and using our theory we can estimate fitness VA, and much more!
I got some needed rest, collected my thoughts, and am here once again here to reiterate something: I think malicious personal attacks within the science Twitter community are wrong, and should not be tolerated. We owe each other better treatment than this, especially these days.