Hi Pablo,
Can I randomly select 20 genes and then choose the best one based on the predicted position?
Thanks,
Shengshuo
Created by Shengshuo Huang shuangat Hi zincfist,
from the beginning we said that we might change the scoring strategy, we hope we will not.
You can do whatever you want (even non-scientific) as long as we understand what you did and you did not use the localization of all 84 genes, but of only the required 20/60/80 subset.
tx
Pablo I presume the rules as stated in the original description will be honored without adding or modifying the allowed strategies. Will it not be acceptable that we treat 84 driver Gene based predictions as gold standard and try to reproduce these outcomes from a subset generated through a scientific approach? Hi, yes Adi, thanks for clarifying, you are correct.
I meant that in the context of someone who wants to have a Data-science-only approach, you don't even need to use the sub-selected driver-genes in situ information and can only use the RNA-seq datapiece and the cells position.
P
Pablo, When you say "you are welcome to not use any of the 84 driver genes and only the RNAseq from the 1297 cells and their position" I am assuming you mean that one could choose not to use any columns in insitu.bin but can use the full RNA-Seq matrix and their position. The way is written sounds that the expression of the 84 insitu genes in the RNAseq matrix is also off limits.
Thanks,
Adi
Well Pareje why do you say that? The whole point of the challenge is not use all the 84 driver genes, if you use them all to select a subset it makes no sense. You can randomly select a subset if you want but you should better select based on some biology, by that I mean using OTHER information from the 84 driver genes related to their position in the regulatory network as indicated in the main text:
Also, you are welcome to not use any of the 84 driver genes and only the RNAseq from the 1297 cells and their position.
Finally DREAM is not merely a data science challenge as we are interested in the biology behind the data. However we try to formulate challenges for data scientists, this is not the exception.
thanks for the interest
So this is not a data science challenge... at all?
(there is a difference between *have to* and *prefer*).
Might have been informative to include this information sometime prior to one day before submission. NO, you have to preselect your genes using some criteria (biological) not based on how well they predict position.