schlosslab / topcuoglu_ml_mbio_2020 Goto Github PK

F1 score
Precision Recall curve AUC

argument with subsamp=F as default

• Argument in function outcome=NULL - defaults to 1st column, otherwise use column name specified
• Make sure conditional outcome is binary
• Factorize outcome column in function

Writing formulas in R:
http://www.win-vector.com/blog/2018/09/r-tip-how-to-pass-a-formula-to-lm/

• documentation that value of 1st sample is 1 and other value is 0
• get names using unique

Right now we call code/R/main.R with 4 positional arguments. Can we change that to flags?

argument with perm=T as default

Hi there,

I'm trying to use the data from this paper for some methods development work, but I'm not sure I understand the metadata properly. The paper states, "n = 490 patients, 261 controls and 229 cases" but I none of the metadata variables break down that way. I looked at the training scripts on this github, and they use dx as the outcome variable, but filter out to only "normal" and "cancer," so it's not 490 patients anymore, and the breakdown of controls and cases aren't 261/229.

Could someone clarify what is the dependent variable these models were trained on?

Thanks!

• For RF: look at Pat's script (lines 46-52): https://github.com/SchlossLab/Sze_SCFACRC_mBio_2019/blob/master/code/rf_classification.R
• LR: hyper-parameters can be passed as argument to function for if we can't figure out a way to automate it
• XGBoost: hyper-parameters can be passed as argument to function for if we can't figure out a way to automate it