Murat Can Çobanoğlu

UTSW Distinguished Fellow. We think drug discovery is a machine learning problem.
Pauca sed matura is a guiding philosophy.

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Cancer drug discovery should be solved as a low rank tensor completion problem: https://t.co/J5VigOhM0H

We describe constructing, using and completing such a tensor; provide true positive and true negative validations.

Thanks to team, incl.: @vasanth1786 @didemagac @gr8straj

[PLOS Comp Biol @PLOSCompBiol]

Have cell types for your single cell data but your visualization doesn't know about them? Check out Haisu. It makes your favorite visualizer (t-SNE, UMAP, & PHATE) label-aware. It also supports hierarchical labels (T cell --> CD8 T cell).
@Cobanoglu_Lab
https://t.co/ZfD58WjQV3

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@oflazer Katılmak isteyen herkesi bekliyoruz!

[Kemal Oflazer @oflazer]

Carnegie Mellon Üniversitesi Mezunlar Derneği Türkiye olarak 19 Haziran 2021 Cumartesi, Türkiye saati ile 16-19 arasında (9am-11am EDT) CMU'lu arkadaşlarımız ile yapay zeka üzerinde konuşacağız. Herkesi davet ediyoruz.

Kayıt Linki: https://t.co/N9wAZ64os7

[Philippe Roudot @PhilippeRoudot]

Another open #PHDPosition in #ComputerVision for #CellBiology. We are looking for an enthusiastic student to push the envelope in the analysis of molecular dynamics across scales and automate navigation in 5D datasets. Info at https://t.co/ymOf1m8sZC. Deadline June 30th!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@HelenGavaghan Yup, exactly :) thanks for reading & sharing our work!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@katrinabmar What I find helpful, when I'm in a similar mental space, is to think back from a positive future. As if you were reading a history book on the topic, in this case the reformation of academia, but written in the future and looking back at today.

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Finally, I'd like to thank my chair @gdanuser1 and my institution @UTSWNews for all the support I've received during the years it took to realize this project.

I hope this tool facilitates cancer drug discovery worldwide. Feel free to reach out with any questions or comments!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Cancer drug discovery should be solved as a low rank tensor completion problem: https://t.co/J5VigOhM0H

We describe constructing, using and completing such a tensor; provide true positive and true negative validations.

Thanks to team, incl.: @vasanth1786 @didemagac @gr8straj

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Predictive engine built using, including but not limited to: python, pandas, numpy, scipy, statsmodels, RDKit, TensorFlow, Tensorly.

Web interface built using nginx, flask, jinja2, Tabulator.

Our code for both parts will also shortly be made available as open source.

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Just listened to @jack on NYT The Daily. One comment stuck: algorithms should be explainable. I think we lack metrics that reward transparency. ML methods (incl in biology) shouldn't compete purely on accuracy, but also on interpretability. Explainable 98% accuracy vs opaque 99%?

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@kD3AN Maybe exams should be different? If a piece of info is trivially accessible, why act like it isn't? When doing a real job, will the evaluation be task completion or internet use to do so? Maybe more focus on analytical ability than memorization?

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@vasanth1786 Thanks Vasanth!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Our new paper just appeared online:
https://t.co/HaeKQOMCWN

Resolving overlap is important for accurate TF activity inference.

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@SpectralFilter @nomad421 @NSF Congratulations Shannon! Great news!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Looking forward to talking at Acıbadem tomorrow!

[Murat Can Çobanoğlu @Cobanoglu_Lab]

James Allison's Nobel prize led us to Stockholm, and the fantastic celebrations here. It is a great joy to attend the events here with my gorgeous fiancee @didemagac !

[Murat Can Çobanoğlu @Cobanoglu_Lab]

@cihancobanoglu Face recognition is quite mature today. Using OpenFace, for instance, one can deploy similar products rather quickly:
https://t.co/9ARNfHn3kO

However, like any new tech, it also presents unique concerns—privacy & security, for instance.

[Murat Can Çobanoğlu @Cobanoglu_Lab]

Demo of our team's deep learning in VR hackathon project—aka deVoReaNN:
https://t.co/xiaGdmYWg3

Repo:
https://t.co/pnijM23dvh

Awesome team; awesome hackathon! :)