Saideep Gona

Former bioinformatician and current PhD student at the University of Chicago. I read the reads. Views are my own.

[Saideep Gona @SaideepGona]

@anshulkundaje And of course those doing the vetting get to live a life of stress and drudgery. 😴

[Saideep Gona @SaideepGona]

@DouglasYaoDY @bfin01322150 @YoAndyZou @ArtirKel This is wrong. Statistical mechanics exists as an attempt to model underlying systems at the individual molecular level with rules. It is incredibly difficult to do at scale so we have to settle for much higher level sampling techniques which average out underlying processes.

[Saideep Gona @SaideepGona]

@ekernf01 @cahanLab @alexisjbattle @josh_weinstock @const_ae @owl_poster Agree on this point 💯. But doesn't this feel like a consortium level effort? I've wondered for a while why benchmark standardization in this and related areas doesn't seem to exist, while there is a huge proliferation of models being developed, each seemingly being the best.

[Saideep Gona @SaideepGona]

@jmschreiber91 Seeing how every new DNA transformer seems to only require 3 months to start to finish, something tells me these aren't isolated results.

[Saideep Gona @SaideepGona]

@BoWang87 Unclear what this has at all to do with AI, just a matter of societal priorities. Seems the real story here is that industry lacks incentive to validate their models.

[Lior Pachter @lpachter]

New paper by @nlapier2 and @hjpimentel on the importance of accounting for isoforms when performing eQTL studies with #RNAseq.
🧬 Isoform-aware methods have greater power than gene-level methods
🧬 Not accounting for isoforms can inflate false discoveries
https://t.co/QqCJR3LvM1

[Siming Zhao @SimingZhao06]

Glad to see our paper, with @xinhe2 @mstephens999 @wescrouse Sheng and Kevin, is out today at @NatureGenet! https://t.co/hMfTQYsGuK. We developed a new statistical method, causal-TWAS or cTWAS, to reliably identify causal genes in genome-wide association studies (GWAS). 1/n

[Saideep Gona @SaideepGona]

@temicrates @AP Temi rhymes with sexy

[Hae Kyung Im 任慧耕 임혜경 @hakyim]

On the problem of inflation in TWAS

Does TWAS yield inflated p-values?

[Saideep Gona @SaideepGona]

@marktenenholtz Matlab makes perfect sense for learning machine learning from scratch. Matrices and matrix operations are intuitive, letting you focus on the actual math. It's accessible to people who don't have python experience. Or is the point that learning fundamentals is a waste?

[Saideep Gona @SaideepGona]

@lpachter Point taken 😂

[Saideep Gona @SaideepGona]

@tangming2005 The way I *feel* it should be is: bioinformatics -> information processing such as sequence alignment, databases, etc. Compbio -> higher level modeling, in-silico experiments, etc. Lots of gray areas of course. Compbio built on robust bioinformatics

[Mark Tenenholtz @marktenenholtz]

Python is removing the GIL.

The GIL (Global Interpreter Lock) prevents you from running multi-threaded code.

That makes ML code, in particular, really hard to write in pure Python.

Here's what it takes to remove the GIL:

[John Novembre @jnovembre]

Here's a new explainer on genetic vs genealogical ancestry that I hope many find useful. https://t.co/FSIwPOTQtU. It's an outgrowth of the recent genomics & population descriptors report and made possible by the great support staff and graphics teams working with @theNASEM

[Saideep Gona @SaideepGona]

@simocristea Depends how you define "lots". A cell with CD4 will often be measured to have no CD4 with some probability. The naive approach would classify those as false negatives, which is wrong. This actually has little to do with biological logic, unfortunately.

[Saideep Gona @SaideepGona]

@temicrates Almost

[Saideep Gona @SaideepGona]

@temicrates Group project

[Saideep Gona @SaideepGona]

@Fox_the_Old @cHHillee I think he's saying it just performs badly by default, but is rescued on some problems

[Mike White @genologos]

Our new preprint: Using synthetic biology we answer the question, why does one transcription factor binding site act differently in different contexts? Check out our surprisingly simple model of regulatory DNA.

Great work by Kai Lowell and @rfriedman22!

https://t.co/AqMWvI1AKB