Results not Matching

[SayanNanda]

Hi,
In order to verify the working of the program, I tried running the script on data for which we already have the results. I used the data in the sample found at GSE63473. I ran the scripts and was able to produce the digital expression matrix.
However the values seem to deviate. Even though the ratios in general hold, when I look at the exact values there appears to be a large difference. For example, for the cell ATCAGGGACAGA there is a very large difference if you calculate the squared difference for each gene.
Is there any explanation for this difference? Is this because the digital expression matrix (in the supplementary files) was created using different versions and a different pipeline. If this is the case, is it possible for you to direct me to source for which I am able to test knowing the results.
Thank you,
Sayan

[alecw]

Hi @SayanNanda ,

Yes, current version of our software does not reproduce old results exactly. If you need to reproduce exactly, you can download the old version here.

Regards, Alec

[SayanNanda]

Hi Alec,
Thanks for the reply.
Are there any examples (which you provide or something that is publicly available) of the input files with the digital expression matrix already created which I can use for the newer version?
Thank you,
Sayan

[jamesnemesh]

We haven’t published additional data with the new method. We do have a few papers in the manuscript writing stage, so there will eventually be some additional DGE data using the most up to date pipelines. You may have noticed some differences in some genes, but if you check the correlation of expression you should see correlation is quite high (I think it’s something like cor>0.99.). There are a handful of genes with fairly different results where the gene boundaries may not be correct in the metadata, and we’re more careful in how we handle those bounds now - reads are contained inside bounds instead of overlapping bounds.

…

-Jim

On Jul 18, 2019, at 12:49 PM, SayanNanda ***@***.***> wrote: Hi Alec, Thanks for the reply. Are there any examples (which you provide or something that is publicly available) of the input files with the digital expression matrix already created which I can use for the newer version? Thank you, Sayan — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#136?email_source=notifications&email_token=ABCZXJYPUWUHJZOLK57AFGDQACNI7A5CNFSM4IET3MK2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2JDGXA#issuecomment-512897884>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABCZXJ7E7FG6ORA3DLXUES3QACNI7ANCNFSM4IET3MKQ>.

[SayanNanda]

Thank you, this was really helpful!