ding-lab/msisensor

Tumor-only method

EgirardBioinfo opened this issue · 23 comments

Hi,
Thank you for your improvment for tumor-only analysis. Could you give us more informations about the method used please ?
Bests,
Elodie

Hi,
When giving the tumor only data, we use the information entropy theory to perform microsatellite instability analysis. The details of this method are described in the paper we are preparing.
Thank you for your support of this software.

I am also very curious about the algorithm behind tumor only mode, esp., how to generate a baseline (using information entropy theory?). Could you share a preprint if available? Thanks

The manuscript is being prepared and we will deliver an uniform cutoff.

Hey!
The output of a tumor only case is 21748 3148 15.01. I assume that 15.01 is the msi_score. Is this score msi-high or msi-low? How do I get the status and what is the cutoff value?
Thanks in advance.

Could you list out some resource on understanding what is the comentropy theory? Majority of the results are masked by entropy and information gain.

Hi
I have checked the MSI status using normal- tumor data, in that case the percentage of MSI is showing very low, to counter check I have analyzed the same data in tumor only, but to my surprise the percentage of MSI sites is significantly high on comparison with the previous one. Could u please tell me how this differences arise? Also I would be glad to know the accuracy and performance of this tool.

Hey! The current tumor only option based on entropy theory is just a PRE-beta release. If you just want to detect MSI status for tumor only specimen, please provide me an email, and I will send your msisensor2 to test your cases.

Is msisensor2 different from version 0.6? If so, could you send me msisensor2 as well to dauss75@gmail.com?

We have quite a bit of data for normal-tumor vs tumor only and trying to detect MSI for tumor only.

Thanks

Hi @Beifang,

My email address is erprateek.vit@gmail.com. Could you send the msisensor2 resources so we can experiment? Appreciate your response!

Hi @Beifang,
I didnt receive the MSIsensor2 till now. Could u please send me to this mail id-uvamivi@gmail.com. Thanks much.

Hey, since many people asked for msisensor2, we are considering how we should provide that to the community.

Please try MSIsensor2 here: http://niulab.scgrid.cn/msisensor2/

Thanks for the link. Could you kindly let me know how msisensor2 is different from msisensor0.6?

The tumor/normal pair model in MSIsensor2 is the same as MSIsensor0.6, but we are using a new tumor only model in MSIsensor2 instead of premature entropy method in MSIsensor0.6. We are going to remove tumor only option in MSIsensor0.6, and only keep tumor/normal pair model in MSIsenor0.6. So, if you have tumor only data, please use MSIsensor2 for MSI status detection.

Thanks.

Any plan to add hg19 as well?

FYI, while installing on CentOS 7, one dependency problem with glibc has occurred.

Could show more details about the dependency problem.

Please see the details below.

$ ./msisensor2
./msisensor2: /usr/lib64/libm.so.6: version GLIBC_2.23' not found (required by ./msisensor2) ./msisensor2: /usr/lib64/libstdc++.so.6: version GLIBCXX_3.4.20' not found (required by ./msisensor2)
./msisensor2: /usr/lib64/libstdc++.so.6: version CXXABI_1.3.8' not found (required by ./msisensor2) ./msisensor2: /usr/lib64/libstdc++.so.6: version GLIBCXX_3.4.21' not found (required by ./msisensor2)

Regarding your problem, we have repackaged MSIsensor2 and updated it on the website.Tests on many servers have indicated that this problem does not exist in the current version. You can download and unzip it again. If you have any problem, please feel free to contact us.

Try download MSIsensor2 again, and use "tar -zxvf" to unzip it. If you still have the problem, you can show me more details about the error.

Use command ./msisensor2 msi –M models_hg38_1000genes -t Sample.tumor.bam -o output.tumor.prefix. If you still have any problem, please feel free to contact us.

Hi
Thanks a lot. It is working perfectly now.

Regards,
Yuvarani