Saulius Lukauskas

Data Science Manager @ GetYourGuide

[Charlie Harris @charlieharris01]

Unexpected development!!

[Saulius Lukauskas @lukauskas_]

@ATvardovskiy @BB_UPenn @PennEpiInst Congrats Andrey! So happy to see this dream of yours come true! :)

[Andrey Tvardovskiy @ATvardovskiy]

Excited to see our paper "Decoding Chromatin States by Proteomic Profiling of Nucleosome Readers" out in @Nature!
https://t.co/vqva7OikoA
Explore our data on how epigenetic modifications modulate protein binding to chromatin with MARCS https://t.co/b5FIzQYHNm
Details below 👇

[Joe Greener @jgreener64]

Really excited to be starting a group among such amazing scientists.

I will be advertising for postdoc and student positions in the next few months. Feel free to email me to discuss informally.

[Andrius Kubilius @KubiliusA]

It took 3️⃣ days for Lithuanians to collect 5 mln euros to buy #Bayraktar for #Ukraine. Proud to be Lithuanian!
🇱🇹🤝🇺🇦

Hope to see Germans and French people to do the same. It’s more important than Scholz and Macron never ending phone calls to Putin!

https://t.co/a7INKx5YjJ

[Saulius Lukauskas @lukauskas_]

@AnnaNieborak I'm certain it's Dr. Nieborak

[Saulius Lukauskas @lukauskas_]

Fantastic thread

[Jacob Schreiber @jmschreiber91]

Found out last night that @NumFOCUS funded my proposal to rewrite #pomegranate from the ground up using @PyTorch as the backend! Need to train massive HMMs using multiple GPUs, or want a mixture of negative binomials as part of your neural network? Watch this space!

[Saulius Lukauskas @lukauskas_]

@nilshomer There is little biological significance in statistical significance. See also https://t.co/QRfFopeQEh

[Saulius Lukauskas @lukauskas_]

@nilshomer If there was one course that I could recommend to everyone, junior or senior, it's @rlmcelreath's "Statistical Rethinking" https://t.co/aMmfld9sq6

[Saulius Lukauskas @lukauskas_]

@nilshomer Be aware that some issues in data are easier to fix in the lab than in silico.

[Saulius Lukauskas @lukauskas_]

@jeffvierstra https://t.co/wcvPdYdHJi

[Oded Rechavi @OdedRechavi]

Not too many people know Sinatra wrote My Way about a bioinformatician

[Robert Schneider @robert_ife]

Please RT. We’re looking for a Computational Biologist in the Schneider lab. #Epigenetics. Integration of omics datasets, #singlecell, embedded in wet-lab + great ML and AI expert environment

➡️ https://t.co/ImmEevUfb0

[Saulius Lukauskas @lukauskas_]

Is accurate software maintained longer, or is well-maintained software more accurate? Either way, I wish there were more ways to thank those who maintain their software post-publication.

[Stephen Turner 🦋 @stephenturner.us @strnr]

Sustained software development, not number of citations or journal choice, is indicative of accurate bioinformatic software https://t.co/02kbhZk5WW

[Kevin Patrick Murphy @sirbayes]

I am delighted to announce that my new book, “Probabilistic Machine Learning: An Introduction”, is finally available in print format! You can order it from https://t.co/fx92WBQvk3, or from Amazon. Also available at https://t.co/dSlKkwYpLr 1/4

[Saulius Lukauskas @lukauskas_]

I vouch for this as well. Maybe not the number of data rows, but more important model assumptions (e.g. that there are no duplicates in data) should always be encoded in assertions. In general, fail early, fail loudly is almost always the way to go.

[Bad Data Scientist @baddatasciencer]

Stupid little data analysis tip: I used to write comments like
"# num rows = 100" as I was exploring data.

Now I usually change that to "assert len(df) == 100"

This way the code will let me know when/if that assumption needs to change, instead of relying on me checking later

[Saulius Lukauskas @lukauskas_]

I like where @uniprot is going with the new beta https://t.co/jqNG2Ek9ce, though I find the fonts and information more legible after zooming out by 10%.

[Saulius Lukauskas @lukauskas_]

This looks like it could be a massive time saver in EDA step if it works as well as advertised.

[Till Nagel @tillnm]

Kick-start your EDA! @lux_api helps visually exploring your @pandas_dev dataframes and renders simple visualizations to view correlation, distribution, and more in a notebook environment. Stumbled upon it just now and find it really useful. https://t.co/USML2Bhlip

[Will Kurt @willkurt]

Probably obvious to many people, but this morning I just realized that R^2 (coefficient of determination) can be viewed as 1 - MSE/Var(y).

For whatever reason my brain puts SS_res and MSE in totally different places where they're very nearly the same.