FAQ  •  Register  •  Login

2017-Shaham et al-Bioinformatics

<<

mleipold

Guru

Posts: 5793

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Mon Apr 24, 2017 3:51 pm

2017-Shaham et al-Bioinformatics

"Removal of Batch Effects using Distribution-Matching Residual Networks"
Shaham U, Stanton KP, Zhao J, Li H, Raddassi K, Montgomery R, Kluger Y
Bioinformatics, 2017
http://dx.doi.org/10.1093/bioinformatics/btx196
<<

mleipold

Guru

Posts: 5793

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Tue Apr 25, 2017 7:08 pm

Re: 2017-Shaham et al-Bioinformatics

Interesting work, especially with the ultimate goal of this in the Conclusions:
"We are currently developing applications of MMD-ResNets to perform calibration in scenarios where multiple batches are present, each batch contains multiple samples and all batches contain a reference sample."
- That's basically how we run all of our customer assays, with an in-house control on every plate.
- Are you planning to implement it on a stim-by-stim basis (in your paper, US and PMA+I), or are you planning to consider only all of a particular control (including all stim conditions within)?


One question, though: I'm not sure what is meant by this comment on page 4, section 4.2.2:
"A typical CyTOF sample contains large proportions of zero values (up to 40% sometimes) which occur due to instabilities of the CyTOF instrument and usually do not reflect biological phenomenon."

1. Do you mean zeroes, as in cells that biologically don't express a marker (eg, CD19 on T cells)?
2. What instabilities of the instrument are you referring to? I mean, the CyTOF does have some instabilities (tuning drift over long run-times, etc), but I'm not sure that's what you mean in this context.

Return to Literature

Who is online

Users browsing this forum: Bing [Bot], Google [Bot], mleipold and 50 guests

cron