uhh, anaconda just sent a message to our HPC admins that we're in violation their ToS and we now need to pay for a license or remove all their software from our system?
I wonder if middle authorship, when you really care about the paper, is like being a grandparent: you get to do a bunch of the fun parts, and first and last author parents have to clean up all the shit.
I feel like bioinformatics and bike maintenance have a lot in common: a multitude of standards, impenetrable documentation, weird multifactorial dependencies, everything just generally being dirty and fiddly, but from the outside it all seems like it should be so simple.
somehow I think the people hyping containerized workflows as a way to enable reproducibility did not anticipate the side effect "now everyone can just run pipelines they don't understand full of software they don't understand to get results they don't understand."
I cannot imagine anaconda/miniconda could have become so ensconced in academic software development if people had not believed it was/would be free, at least for academic use.
8yo tells 6yo that Dimetrodon is not a dinosaur. 6yo pitches a fit, screams that it is. Daddy tries to teach them both tree thinking, inadvertently solving the dispute by boring them so much they just want to get back to playing.
it seems totally bonkers to release an apparently free package installer that provides users and admins no warning that a paid license is required, wait YEARS, and then come knocking and demanding to be paid.
As part of a two-body problem that did not get resolved the way we wanted, I do not think this article is helpful or remotely reflects the barriers we faced.
It seems like there should be a german word for the feeling you get when you've searched google scholar on a topic 100 times and then somebody tells you about a paper with a million citations and all your key words in the title and you somehow missed it.
I genuinely can't believe that someone had to write this paper in 2024. Any application of SNP data that depends on accurately identifying heterozygous genotypes is going to be STRONGLY INFLUENCED by depth of coverage.
A must read for anyone studying population genomics.
Low-coverage sequencing and Wahlund effect severely bias estimates of inbreeding, heterozygosity and effective population size in North American wolves
1980s: here: admire my fine table of Fst values.
1990s: look at this gene tree! so much better than stupid old Fst!
2000s: that tree is lying. to. you. fit this model instead!
2010s: ok I have a genome and your stupid model won't even run I'm just going to calculate Fst.
2019:
Congratulations to
@jeromekelleher
,
@gil_mcvean
,
@leo_speidel
, and
@simon_r_myers
on publishing their fabulous new methods for inferring genealogical trees from whole genomes! Here's my take on why these papers make it a great day for population genetics:
I'm curious about examples of strong genetic differentiation in eukaryotes on very small scales. I'm helping out on a study of a desert plant species where populations 250m apart separate out cleanly in a PCA. it kind of blows my mind.
is this a hot-take?
when you need a phylogeny, sometimes a neighbor-joining tree on uncorrected distances is good enough to make a point and a more intensive analysis is pointless extra effort.
Unpopular opinion: virtually all words and phrases in this thread are fine and banning them would do literally nothing to improve the readability of scientific literature.
When you finish a PhD in phylogenetics they try to take you to a special room to tell you something but there are just so many rooms they can't guarantee they've taken you to the right one, so they keep trying new rooms and eventually they just give up and tell you to postdoc.
When you finish a PhD in theoretical ecology, they take you to a small room and tell you all models are correct and most are useless. I’m posting this at great metaphysical risk to myself and my family.
if machine learning can't tell me who wrote the anonymous trump op-ed then I don't want to hear another damn word about it solving all of evolutionary biology.
I'm interested in recent books on genome biology/evolution along the lines of Lynch's Origins of Genome Architecture, which is now 15 years old. Any suggestions?
I would love to know how many gene-phenotype or gene-environment associations in conservation genetics and molecular ecology stand up to increased sampling, replication, or external validation of some kind.
This R package wants me to run a function that will install software on my system. It has code that literally asks for my password and then runs 'system("sudo -kS sh -c 'make install; ldconfig'", input = password)'
An R function secretly running sudo seems so, so sketchy
@mike_c_mahoney
... is destructive. I wish I knew how I came around to it, I think we all know that simple facts rarely convince people with strong feelings, but I'll never have an outdoor cat again.
I have long gotten misdirected e-mail intended for a certain beloved Canadian actor from Schitt's Creek. Today a guy asked me to make a video wishing his girlfriend, who is a huge fan, a happy birthday. Thinking about doing it.
fascinated by the gay genes study? this is a great time to read (or reread) this blog post by
@Graham_Coop
on the complexities of understanding the genetics of human traits. it's largely focused on btwn population diffs, but still totally relevant.
throwing my 2c in on the retraction scandal:
1) twitter *can* be horrible, but the response to this I am seeing is so generally mature and supportive of the folks who are hurt by it, that I am encouraged.
I don't know who needs to hear this, but good editing is so much more than ruthlessly scratching out "elucidate" and "in order to". In fact, I think attending to those is a complete distraction from an editor's real task of helping to organize and clarify ideas.
Is it too early to plug your stuff for
#Evol2019
?
I'm giving a poster on Sunday on a cool project I've been working on: sex chromosome evolution and sex determination in two species of killifish.
I'm glad I've restricted most of my twitter follows to science. I keep getting sorely tempted to come here and rage about politics, which would not help anybody, but every time I do I get distracted by a cool paper.
It seems like there are direct estimates of mutation rates being published for many organisms, often buried in larger manuscripts. Anybody want to start putting together a google sheet to track them? Is this a silly idea?
A publicly editable google sheet initialized with my two current favorite mutation rates. I'm not sure which fields would be most useful, or which methodological details are most pertinent. I'm open to suggestions about better ways to do this.
It seems like there are direct estimates of mutation rates being published for many organisms, often buried in larger manuscripts. Anybody want to start putting together a google sheet to track them? Is this a silly idea?
properly simulated data will be MUCH harder to detect. and every year, mountains of graduate students in a very competitive environment are developing exactly the necessary skills. I don't know what to say except that this really disturbs me.
more digging into other people's R code... these spanish coders are using data frames named "datos" and "datitos". gonna have to steal that convention.
Well now I want to know the minimum number of individuals in the US who have to use a genetic ancestry site before no criminal who leaves DNA behind can escape detection. I'm sure that's a blog post
@Graham_Coop
probably doesn't want to write .
@FrankBoerman
I am not an expert in package management, but we don't just use conda for python packages, and frequently our users are simply following suggestions provided by software developers that they can either spend days in dependency hell or use conda.
I resolved a little RADseq mystery... we have this dataset where there is bimodal coverage of rad sites. there is a wide, high coverage mode at ~ 100x, and a narrow low mode at ~7x (marginal Y-axis histogram).
2) if this is indeed fraud, the only reason it was caught is that the responsible person didn't know how to simulate data. next to becoming a credible expert in a scientific field, getting an academic job, and publishing loads of papers, learning to simulate data in R is TRIVIAL.
What they claim: simply choosing the most parameterized phylogenetic model available is not risky enough to justify making everyone do intensive model selection.
What I think everyone has heard: models don't matter in phylogenetics. wheeeee!
I just submitted ~600 17 year old chipmunk Sanger sequences to genbank for a manuscript I thought was long dead, but somehow came back to life and is now accepted for publication... pending accession numbers.
Has anyone ever done a study on the extent to which savings from bulk purchases at places like costco are offset by increased consumption he thought to himself as he ate 12 strips of dried mango.
@mike_c_mahoney
I felt terrible when he killed things but when somebody first told me about the toll outdoor cats take on wildlife, like tons of other people I got angry and defensive and rejected it. It sucks when somebody tells you that something you love, and I did love having my cat outdoors
It's amazing that biology can be done at this taxonomic and geographic scale now, but I wonder how much credit the 13 museums listed in the acknowledgements of this paper will get for making a paper like this, with only 5 authors, possible.
I can't get over that he opens with that shit about delivering babies. personally I don't think anyone deserves the title doctor unless they have stitched together parts of various corpses into a hideous monster and then successfully brought it to life with a bolt of lightning.
I love birding, but man I question my choices when I'm thinking about taking some of my precious free time to go to a nail salon parking lot to look at dirty gulls lounging on the roof of a waste transfer station.
I really enjoyed the ASN presidential address by
@MikeWhitlock63
on scientific methodology. I think his intro on the absurdity of "p < 1e-253" should probably be a shot across the bow for all of us doing genomics stuff.
#Evol2019
I really like this review. It is extremely accessible for a difficult (to me anyway) topic:
Patten, Manus M. "Selfish X chromosomes and speciation." Molecular ecology 27.19 (2018): 3772-3782.
"What do you mean you've never heard of the gene ND5!? It was only the third most popular gene used for single-locus mtDNA phylogeography in the 2000s!"
I'm not a fan of the commonly seen numerical system for denoting bayes factors.
Let's embrace 21st century symbology and do better:
k < 1🤔
k = 1🤔
k > 1🤔
k > 10🤔
k > 100🤔
I’m not a big fan of the commonly seen *, **, *** system for denoting p-values on graphs.
Let’s embrace 21st century symbology and do better:
p < 0.1 🤥
p < 0.05 🤔
p < 0.01 🙂
p < 0.001 🤩
p < 0.00000001 🤯
I'm feeling completely overwhelmed by all the stuff I have to do, so naturally instead of doing any of it I lubricated my squeaky office chair. It was such a glorious success I'm calling it a day.
Trivial coalescent/variant calling question: GATK docs state "To minimize the chance of missing singletons, we increase the cohort size -- so that singletons themselves have less chance of happening in the first place." But this is false, right?
if you're in northeast connecticut, or maybe anywhere in southern new england and you like birds even a little, you should go birding right now. migration over night was exceptional.
Maybe I'm just not up on the methods, but I don't understand how the "model-based demographic inference" and the for lack of a better term "oh my selection really seems to very strongly shape genetic diversity" subfields of population genetics coexist.
the three phases of statistics whiplash:
1. this impenetrable jargon seems impossibly complicated.
2. ha! this technique is actually stupidly simple! I understand it!
3. it turns out the actual justification for it is 5 solid pages of math that I don't understand at all.
Top Mistakes by which PopGen nerds give themselves away all involve parameter names:
1. "No, Ne has nothing to do with the number of individuals in a population."
2. "You're a fool if you interpret 'admixture proportion' as an admixture proportion.
3. "Linkage disequilibrium"
The top mistakes by which non-pop-gen folks give themselves away all involve Hardy-Weinberg:
1. Fail to check HW
2. Assume HW violations are biologically meaningful rather than genotype errors
3. Even w/o HW violations, analyze genotype frequencies rather than allele frequencies
Question for my systematics colleagues: at this point in science history, when if ever is it permissible to publish a morphology-based phylogeny (or character-mapping) study using only parsimony not maximum likelihood/similar methods? (can of worms I know)
@mike_c_mahoney
I grew up with an outdoor cat that I loved. He was calm and affectionate. All the indoor cats I knew were weird and squirrelly and even worse, used nasty litter boxes. I loved watching him prowl outside, and also loved wondering what the hell he was doing out all night.
@Graham_Coop
you're misinterpreting. this is obviously a circular cladogram showing that earth is the ancestor of the rest of the planets, and that gas gigantism has evolved multiple times.
Population geneticists in publications: look at all the cool things you can infer from the SFS!
Population geneticists informally to each other: what the hell are all these snps doing in these bins
Many of the population geneticists I follow here have openly discussed struggling with factors of two. Have any of you considered reaching out the IHME to counsel them in these trying times?
@LAevolving
This is almost certainly because there is a young-ish, non-degenerate sex chromosome with a very large non-recombining region. Divergence between the still largely homologous X and Y (or Z/W) that exceeds population structure drives the signal.
so what's the bioinformatic model for population genomics with assembled whole genome sequences (vs reference mapping, vcf, etc)? who's figuring that mess out?
@ethanblinck
"you see those birds out the window? from a, uh, genetic, uh, perspective, there, uh, probably aren't as many as it looks like there are? what I mean to say is, uh, they've probably got fewer snps than you might guess"
Bioinformaticians: Ugh. Noobs use horrible file naming conventions that increase the likelihood of errors and hamper reproducibility.
Also bioinformaticians: please update to the latest version of my software package, 3.1.10, which succeeds 3.1.9.
Do frogs have weird 3' UTRs? Like generally huge ones? I'm looking at a gene expression dataset for a frog in IGV and every gene has like 4-15kb of coverage at the 3' end. I don't think I've noticed this before.
@amanda_dettmer
unsolicited recommendation: our dog takes bravecto, and she hasn't had a single tick since. we started b/c she got lyme and anaplasmosis within a month of moving to connecticut.
@PartridgeCG
I think it's supposed to give a candidate a chance to completely collapse from pressure and exhaustion. Or at least, I used it for that once.
A 16 hour rollercoaster:
Up: new data! spectacular finding! what does it mean??
Cresting: wait, huh, that can't be right
Down: OH GOD SAMPLE LABELS ARE SCRAMBLED
Bottom: I can't afford to redo this experiment
Up: A clear paper trail allows for label deconvolution.
...
fun fact I just learned about Stacks: the VCF output when using reference mapped data doesn't put the reference allele in REF and the alternate allele in ALT.