RE: Inquiries into software Canopy

[Chen, Hui-Zi]

Hi Dr. Jiang,

 

First of all happy holidays to you!

 

Thank you again for your assistance before with CANOPY. Our bioinformatics group have come up with a few questions in our genomics analysis of SCLC that we hope you or your colleagues may help answer. Please see below…

 

We inspected matrix C from your provided MDA231 test data, and we were unsure how it correlates to other matrices. Where are labels such as “chr7_1” referenced in the data set, and how should overlapping CNA events be named? In addition, WM and Wm appear to reference whole chromosomes, rather than labeled CNA events within them. After manual inspection, how should overlapping CNA events be incorporated into Canopy input? Thank you!

 

Warm regards,

Huizi

 

From: Chen, Hui-Zi
Sent: Friday, November 17, 2017 3:46 PM
To: 'Jiang, Yuchao' <yuc...@email.unc.edu>
Cc: Bonneville, Russell <Russell.B...@osumc.edu>; Roychowdhury, Sameek <Sameek.Ro...@osumc.edu>; Wing, Michele <Michel...@osumc.edu>; Miya, Jharna <Jharn...@osumc.edu>; canopy_p...@googlegroups.com; Zhang, Nancy R <n...@wharton.upenn.edu>
Subject: RE: Inquiries into software Canopy

 

Hi Dr. Jiang,

 

Thank you for your quick reply and helpful hints. We will meet as a group to go over the next steps in our analyses. I have a feeling we may need additional help and suggestions in the future and will be sure to reach out at that time.

 

We thank you again and have a wonderful weekend!

Huizi

 

From: Jiang, Yuchao [mailto:yuc...@email.unc.edu]
Sent: Wednesday, November 15, 2017 2:38 PM
To: Chen, Hui-Zi <Hui-Z...@osumc.edu>
Cc: Bonneville, Russell <Russell.B...@osumc.edu>; Roychowdhury, Sameek <Sameek.Ro...@osumc.edu>; Wing, Michele <Michel...@osumc.edu>; Miya, Jharna <Jharn...@osumc.edu>; canopy_p...@googlegroups.com; Zhang, Nancy R <n...@wharton.upenn.edu>
Subject: RE: Inquiries into software Canopy

 

HI Huizi,

 

Thanks for your interest in Canopy! For CNA in cancer, especially overlapping CNAs, it is non-trivial to call accurately and this is important for Canopy in that otherwise it would be garbage in garbage out. Have you looked at the pages below?

 

https://github.com/yuchaojiang/Canopy/blob/master/instruction/SNA_CNA_input.md This is how to generate CNAs by FALCON and Sequenza. I believe there is QC included in the demo code by FALCON to curate the segmentation results although it is recommended that you also examine the results to make sure it makes sense.

                                                                                                                                                                                                                                           

https://github.com/yuchaojiang/Canopy/blob/master/instruction/SNA_CNA_choice.md  on how to choose CNAs (basically, it is recommended to load FALCON or FALCON-X’s segmentation results into IGV for visualization and focus on the large (sometimes even chromosomal arm level) CNAs and the CNAs that show distinct patterns between the samples from the same patient).

 

https://github.com/yuchaojiang/Canopy/blob/master/instruction/overlapping_CNA.md  This is how to assign values in the C matrix for overlapping CNAs. In this case, you need to visualize the segmentation results otherwise it wouldn’t be identifiable (i.e., a one-copy loss can be a one-copy gain followed by a two-copy loss –the latter (two overlapping events) is mathematically not separable from the former (one event) plus biologically it is rare too). Therefore, at this stage, this still requires a lot of manual picking and usually this works better than automation. For example, a homozygous deletion within a heterozygous deletion will give you four breakpoints and three levels of normalized results (normal, 1-copy loss, and 2-copy loss). In this case, you know there are two CNV events.

 

I hope this address some of your questions. Feel free to let us know if you have more!

 

Yuchao

 

 

From: Chen, Hui-Zi [mailto:Hui-Z...@osumc.edu]
Sent: Wednesday, November 15, 2017 2:07 PM
To: Jiang, Yuchao <yuc...@email.unc.edu>
Cc: Bonneville, Russell <Russell.B...@osumc.edu>; Roychowdhury, Sameek <Sameek.Ro...@osumc.edu>; Wing, Michele <Michel...@osumc.edu>; Miya, Jharna <Jharn...@osumc.edu>
Subject: Inquiries into software Canopy

 

Dear Dr. Jiang,

 

I am a senior medical oncology fellow at the Ohio State University Comprehensive Cancer Center working with Dr. Sameek Roychowdhury (https://precisioncancermedicine.osu.edu/meet-team).

 

My research currently focuses on characterizing tumor heterogeneity in relapsed or platinum-refractory small cell lung cancer through whole exome sequencing of metastatic tumor samples procured from research autopsy. My work involves close collaboration with our bioinformatics group (Russell and Jharna, copied here).

 

In our work with SCLC, we have applied the software tool you developed, Canopy, to our WES data in order to infer subclonal structure and cancer phylogeny. We have successfully run Canopy with SNA input from our samples, but have not yet been able to incorporate CNA input. We are currently using FALCON to generate allele-specific CNA calls. We are hoping and would like to learn more of QC procedures and CNA curation, along with methods to distinguish overlapping CNA events from shared CNAs (to generate matrix C). Are there any guides, how-tos, best practices, or more generalizable sample code we could consult?

 

We thank you very much for your time and help with our inquiries.

 

Looking forward to hearing from you soon.

Warm regards,

Huizi

 

 

Huizi Chen, MD PhD

Medical Oncology Fellow

Divisions of Hematology & Oncology

Department of Internal Medicine

Comprehensive Cancer Center

The Ohio State University Wexner Medical Center

Email: Hui-Z...@osumc.edu

Phone: 614-477-2678

[Jiang, Yuchao]

Hi Huizi,

How to deal with overlapping CNA events is non-trivial and at this stage it still needs a lot of manual inspection and selection. Let me try to answer your question here and let me know if you need more clarification:

For the MDA231 dataset, we know chr7 has two CNA events (supplementary Figure S13 in our paper) with different breakpoints. However for this CNA region shown in Figure S13, we only have one CNA input (from WM and Wm) for a specific sample, i.e., we won’t know the chr7_1 and chr7_2 input for each sample until we perform the deconvolution. Therefore, we name them chr7_1 and chr7_2 just so we know that there are two CNA events on chr7 that are overlapping, as specified by the C matrix.

For WM and Wm matrix, there will be only one major copy and one minor copy for one genomic segment that harbors any CNA (it’s not a value per chromosome but per copy number segment). However, multiple CNA events can overlap with this genomic segment/region. In matrix C, each row is one copy number REGION and each column is one copy number EVENT. You can only correctly assign this if you manually visualize and inspect the data otherwise it won’t be identifiable. For example, a homozygous deletion can always be treated as two consecutive heterozygous deletion but in this case we wouldn’t separate these as two events. However, for the case below (three samples from patient GBM9, with ASCN results loaded into IGV for visualization), for example, we know that there is a duplication, followed by a loss-of heterozygosity (and thus two copy number EVENTS) in the yellow copy number REGION.

I hope this helps!

Happy holidays!

Yuchao

[Roychowdhury, Sameek]

Thanks very much for your time and feedback!

 

Sameek