Thread by @lpachter

A thread on how to analyze *any* single-cell RNA-seq dataset *without a computer* !! This 🤯 magic made possible thanks to insights of @vntranos, @sinabooeshaghi and @pmelsted + work on kallisto | bustools 🐻🚌🛠️ by the authors of biorxiv.org/content/10.110… 1/12

@vntranos @sinabooeshaghi @pmelsted First, a demo: open the website below, click on connect, and then run the @ProjectJupyter notebook on the website by repeatedly pressing shift-enter:
colab.research.google.com/github/BUStool… The notebook should run in less than 10 minutes. 2/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter Congratulations! You’ve just run an analysis of a single-cell RNA-seq dataset from scratch. The notebook downloaded the required software, streamed reads from SRA accession SRR8599150 (mouse retina), and ran kallisto|bustools for you.
ncbi.nlm.nih.gov/sra/?term=SRR8… 3/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter Where did this happen? The notebook ran on @GoogleColab, which is a free Jupyter notebook environment running in the cloud. The notebook was automatically parsed by @GoogleColab from its location in the link (see tweet 1/12). The notebook is here
github.com/BUStools/getti… 4/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab The free environment you used had 12.6Gb RAM, a Xeon single core 2.3Ghz processor & 320Gb of disk. Not a very powerful machine (although @Google does supply a Tesla K80 GPU as well!) but good enough for kallisto | bustools thanks to @pmelsted keeping memory needs <4Gb. 5/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google kallisto | bustools is 250x cheaper than Cell Ranger if run on the @awscloud . But with @googlecolab it's ∞ times cheaper (namely free!), and this is possible as a result of the tiny footprint which was @pmelsted's design axiom throughout the 🐻🚌🛠️ project. 6/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud Another part of the magic is that the sequence reads were streamed to @GoogleColab. See @sinabooeshaghi's post on this for more detail on how this happens. sinabooeshaghi.com/2019/07/09/fas… The example you processed is a small one, but the workflow can be used for any dataset. 7/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud Here is a kallisto | bustools notebook analyzing over a *billion* reads by streaming them directly from the ENA: colab.research.google.com/drive/1vvey6Eh… This takes a few hours but it's ** free **. One. billion. reads. 8/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud The data in this analysis was recently published by Li et al. dev.biologists.org/content/146/12… and our workflow opens it up for many other analyses by anyone. 9/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud For example, with the kallisto | bustools workflow one can output a transcript compatibility count matrix to examine isoform shifts between cell types as demonstrated in this recent paper by @vntranos, @lynn__yi et al. nature.com/articles/s4159… 10/12.

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud @lynn__yi My mind is truly 🤯 Analyzing #scRNAseq has cost labs thousands of $ / month in the cloud just to get count matrices. It’s now free, and the only resource you need is a web browser. We'll be posting @GoogleColab notebooks for other workflows (e.g. RNA velocity) shortly. 11/12

@vntranos @sinabooeshaghi @pmelsted @ProjectJupyter @GoogleColab @Google @awscloud @lynn__yi Special shout out to @vntranos and @sinabooeshaghi who set this up. If you're not following them on @Twitter you should do so immediately. Many more exciting single-cell RNA-seq methods coming from them soon. 🐻🚌🛠️ #scRNAseq #methodsmatter 12/12.