compare the results of DEG analysis with results of v4.0.0
MugenQin opened this issue · 3 comments
Hello CellphoneDB team,
Thanks for your excellent work in providing powerful tools for researchers. The method 3 DEG analysis is useful for us to detect more LR pairs than the previous versions. But I was a little confused while I compared the results of DEG analysis with results from v4.0.0. I expected I can get more LR pairs in DEG analysis method than CellphoneDB v4.0.0 and include the results from v4.0.0. However, the specific LR pairs I detected in CellphoneDB v4.0.0 can't be detected in the results of DEG analysis method. Besides, both the results of v4.0.0 and DEG analysis method are meaningful for us. So could you give me some ideas about this puzzle?
Thanks!
Hi MugenQin,
Thank you for using CellphoneDB and for reaching out to us.
CellphoneDB software package (that you get via pip install) offers three analysis methods: method 1 (basic), method 2 (statistical) and method 3 (DEG) and can be run with the CellphoneDB database versions >= v.4.0.0. Given the above, your statement 'the specific LR pairs I detected in CellphoneDB v4.0.0 can't be detected in the results of DEG analysis method.' is not making sense to me.
Have you run the method 3 (DEG) analysis against the database version v.4.0.0. If so, what exactly are you comparing the LR pairs returned by that analysis to? Many thanks.
Best,
Robert.
Hi MugenQin,
Thank you for using CellphoneDB and for reaching out to us.
CellphoneDB software package (that you get via pip install) offers three analysis methods: method 1 (basic), method 2 (statistical) and method 3 (DEG) and can be run with the CellphoneDB database versions >= v.4.0.0. Given the above, your statement 'the specific LR pairs I detected in CellphoneDB v4.0.0 can't be detected in the results of DEG analysis method.' is not making sense to me.
Have you run the method 3 (DEG) analysis against the database version v.4.0.0. If so, what exactly are you comparing the LR pairs returned by that analysis to? Many thanks.
Best,
Robert.
Hi,
Thanks for your reply. I apologized for my wrong statement. I compared the results of method 3(DEG) with the results of the method based on database version v4(maybe equal to method1?), which I used in the last year. Then I got the result that 'the specific LR pairs I detected in basic method can't be detected in the results of DEG analysis method.'
Qin
Hi Qin,
The method 3 is essentially method 1 in which your DEGs file determines if a given interaction is relevant/significant or not. For a given interaction LR in a cell type pair CT1|CT2 (reported in degs_analysis_means_.txt) file, it is deemed significant (that is: has a non-zero value in degs_analysis_relevant_interactions_.txt and degs_analysis_significant_means_*.txt files) only if in your DEGs file either of L or R are reported as differentially expressed in either CT1 or CT2.
In order to obtain meaningful results from method 3 we advise users to formulate their biological question first, and then run their differential analysis according to that question, and only then use the output of that differential analysis as the DEGs file input to CellphoneDB DEG analysis (method 3).
For the same database version and provided the same call arguments were used for both method 1 and 3, simple_analysis_means_result_.txt file from method 1 should contain the same interactions as degs_analysis_means_.txt from method 3, except that method 1 cannot tell you which interactions are significant and thus is not particularly useful or revealing.
Hope this helps - let me know of any further questions.
Best,
Robert.