Posts on X

Exciting work led by @mlotfollahi and @falexwolf: We show feasibility of in silico predictions of cell-type specific perturbation response, out of sample, across cell types, studies and species. biorxiv.org/content/early/…

Looks indeed ... interesting. What did you do?

@edie_crosse

These #singlecell trajectories sure look festive. Though @fabian_theis I think I may have done something slightly wrong..

Quoted post

Yes, and a single discrete one to have chaotic behavior as well. Not for compartment systems, but surely possible for more complex chemical reaction kinetics.

@RBrinks

@fabian_theis Wohow, a system consisting of 2600 ODEs ... not to mention the many more parameters - most of them estimated. A system of three (!) ODEs with three parameters is enough to impose "chaos" like the Lorenz attractor.

Quoted post

Impressive technical endeavor with inference (albeit rough) of parameters in more than 2600 ODEs - for me a first step towards a ‚big data systems biology‘. t.co/Zn2VS56qJz

Quoted post

Agreed - Happy scientists make better scientists‘!

@OliverStegle

Good points in this. A happy lab is more productive and a fun place to work for everyone. Would add 8. Make sure to keep a spirit of passion for science, driven by curiosity to figure out how things work and not (primarily) papers/deadlines. nature.com/articles/d4158…

Quoted post

That’s soo 2017 - I only dance the U-MAP!

@markowetzlab

@fabian_theis Next class: DANCE THE T-SNE!

Quoted post

Dancing k-means - my new, highly didactic introduction to unsupervised machine learning 🕺

Not sure, but IMHO more if concentrated in a few places than spreading it thin.

Quoted post

I guess most budgets are peanuts vs China’s or Google/Amazon/FB’s - however vs other European & US research budgets I would think it’s not.

Quoted post

Agreed - I find it interesting that CS has done a much better job of owning data science than stats and math.

Quoted post

Agree, Ali - time to catch up; really see this as an opportunity for biomed. Together with @OliverStegle @gagneurlab and others we are currently starting a project boldly called #MechML, where we hope to add some more mechanistic understanding to predictions.

@erturklab

@fabian_theis @faznet It is somewhat late but still timely. We need to scale up efforts in AI quickly, especially to remove the human bias from the biomedical science.

Quoted post

Germany proposes #AI strategy, finally: 3B€ and 100 new professorships. Let’s put many in biomed, where we have actual neural networks as well as large-scale data challenges. faz.net/-ikh-9gmot?GEP… @faznet

Congratulations, Sarah - on the comp side, can see the cellphonedb.org becoming useful for mining cell interactions!

@satijalab

Teichmann lab reconstructs the human maternal–fetal interface with breathtaking resolution to find distinct subsets of perivascular, stromal, and NK cells- and predict cell-cell interactions based on ligand/receptor expression profiles! nature.com/articles/s4158…

Quoted post

Truely pretty - and I don’t even see colors - but now you have to tell us what this is!

Quoted post

Roser @roserventotormo @sangerinstitute presenting her impressive work on fetal maternal interface @TU_Muenchen - gotta love the aptly named CellPhoneDB and the corresponding CellPhoneBook to test for interacting cells!

Interesting read - summary: the “Intelligent Digital Mesh” links people, devices, content and services. And i would love to have one of those AI developers (=an AI developing code for you)!

Quoted post

Thx, Hart, really put some time into the animation there ✨😂

@hartjackson

@andrei_chiffa @fabian_theis “Putting glasses on cells” loved it 🤓

Quoted post

Thx Andrei, also for appreciating our work on the software side - @falexwolf, @trueflyingsheep, @gokcen and @lkmklsmn really put some time into this, and it pays off also in the lab.

@andrei_chiffa

@fabian_theis Followed up with neural networks and deep learning in scRNA-seq: github.com/theislab/DCA. Once again, the library is formatted amazingly.

Quoted post

Inspiring the next new talents - @ICBmunich and @HelmholtzMucEn at the Munich Science Days. Let‘s hope they don’t find out too soon that science is not (only) 3D and pretty pictures.

They had to wear them when building balloons for their loon project 🎈

@JeanetteErdmann

@ceclindgren @saralpulit @fabian_theis What is the point with shark slippers?

Quoted post

Congratulations, Laleh!!

@embl

Laleh Haghverdi (2nd on the right, postdoctoral fellow in the Huber and Marioni groups) was awarded the Research Prize by the Peter and Traudl Engelhorn-Stiftung yesterday, for her pioneering work on single cell transcriptome analysis methodology. Congratulations!

Quoted post

True good point, although the idea of receptive fields (CNNs) have been around for more than 20 years.

@jakhmack

@fabian_theis Sure there is tons of marketing, and many recent successes are a consequence of smart engineering — but some of the key ideas that made it possible (eg CNNs and RL) come from neuro, and it will also be exciting to see where the next generation of ideas will come from...

Quoted post

You are innovative indeed - go for it! @JanHasenauer

@CanPathPro

CanPathPro selected as a finalist for the Innovation Radar Prize 2018!! Please vote for us here: ec.europa.eu/futurium/en/ex…

Quoted post

Congratulations, Alon!

@mpi_psychiatry

Alon Chen elected next President of the Weizmann Institute of Science @WeizmannScience in Rehovot, Israel psych.mpg.de/2407119/pr1811…

Quoted post

At the #sophiasummit on #AI - discussion how much current deep learning research is drawing from neuroscience. Not sure about this hence trying my first poll:

Yeah :) although in this case they needed them to walk on balloon silk they knitted together for their project ‘loon’ 🎈

Quoted post

Yes, she suggested to define criteria when to stop already early on during project design.

@ceclindgren

@saralpulit @fabian_theis That’s what I am guessing too .. I am curious what the speaker said about it ..

Quoted post

That may help

@thalexandrov

@fabian_theis - have $700B market capitalization 😉

Quoted post

Good points: - don’t fear failure - learn fast - tackle the hardest problem 1st - set kill criteria - love problems not tech - contact the real world - interdisciplinary teams - wear shark slippers 🦈

@fabian_theis

@Obifelten talking about how Google X is tackling and often failing moonshot projects at #HelmholtzHorizons - ‘fail gracefully!’

Quoted post

@Obifelten talking about how Google X is tackling and often failing moonshot projects at #HelmholtzHorizons - ‘fail gracefully!’