For a change, a proper #milestone. 1000 reads on RG for the #Reticulate origin of modern plane trees (#Platanus, Platanaceae): A nuclear marker puzzle

https://www.researchgate.net/profile/Guido-Grimm/achievement/6929624c0df56f392494b91f

It's profoundly comforting to see that one's work is still raising interest after more than a decade, especially, something that at the time of publication was harshly attacked by "expert #reviewers" –anonymously of course (#FightTheFog). And ignored by the botanical High Priests beyond the Big Pond 😎😎😎

Took a while, but finally out:

Pruning the Tree: Comparing #OTUs and #ASVs in High-Throughput Sequencing of 5S-IGS Nuclear Ribosomal DNA in #Phylogenetic Studies

How to reduced 100ks of reads down to the 100 really needed, a comparison between #Mothur and #dada2 pipelines using our nigh-#reticulate #beech dataset as test subject

https://doi.org/10.1002/ece3.72242

Na #PythonCerrado2025, tivemos ontem um excelente tutorial do Lucas Marcondes Pavelski https://github.com/lucasmpavelski.

Aprendemos sobre #R, #tidyverse, #reticulate, várias ferramentas essenciais como #ggplot2 e #dplyr, vendo na prática como aplicá-las. Foco na ponte #Python <-> R.

Tudo novidade pra mim, vieram várias ideias interessantes de análises e plots.

#PythonCerrado

holy sh*t can't believe this actually works

xarray$concat(list(vds1, vds2), dim = 'time', coords = 'minimal', compat = 'override')
<xarray.Dataset> Size: 32GB
Dimensions: (time: 2, lat: 17999, lon: 36000)
Coordinates:
time (time) int32 8B ManifestArray<shape=(2,), dtype=int32, ...
...

concat *virtualized* netcdfs with #reticulate #RStats don't think I can parallelize because these are pointers, but could write out the individual Parquet representation I guess

A classic, my first paper on #beech #genetics getting a RG-#milestone: 1000 reads for "The #evolutionary history of #Fagus in western Eurasia: Genes, morphology and the #fossil record" RG added #reticulate #evolution, suppressed #speciation, could say that.

http://dx.doi.org/10.1007/s006060200044

Low impact? Enduring research still (obviously) worth a read over 20 yrs later 😎

PS Still happy with it. But wouldn't just use a tree anymore for #phylogenetics

weird #RStats idea

library(python)

and via #reticulate you just do

somepackage$<things>

on a controlled system, you could set a user config for all the "somepackage"s you want lazy-loaded :)

I have now spent many hours trying to get a particular set of packages to work in #rstats with #reticulate, combining keras, tensorflow and pycox (built on pytorch).
I have tried python 3.8 through 3.10. I have tried mamba instead of conda for dependency solving. I have installed a PPA-based gcc-11 on Ubuntu 20.04 where gcc-10 is the latest. I have installed some intel MKL library that was required even though I'm on an AMD system and don't know why.

Fuck everything about every part of this.