No Instructions for Tissues and Cell Culture Merging

[DarioS]

Would be nice to have advice in the user guide for merging cancer tissue and patient-derived cell-culture. They aren't merging.

[pati-ni]

Please provide a few more details. Does the code throw an error, or harmonized populations are not adequate?

On Fri, Feb 2, 2024, 18:00 Dario Strbenac @.***> wrote:

Would be nice to have advice in the user guide for merging cancer tissue and patient-derived cell-culture. They aren't merging.

— Reply to this email directly, view it on GitHub https://github.com/immunogenomics/harmony/issues/240, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSFW2HFXPNIW7ZVOPO65BLYRVVX3AVCNFSM6AAAAABCXNJDCSVHI2DSMVQWIX3LMV43ASLTON2WKOZSGEYTKOJWGQYDQMI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[DarioS]

The harmonised populations stay separated. There is no error message.

[pati-ni]

Should the tumor cells merge?

What about the rest of cell types, are these separated?

Can you share the objective function plot?

Have you tried higher values of theta?

On Thu, Feb 15, 2024, 04:00 Dario Strbenac @.***> wrote:

The harmonised populations stay separated. There is no error message.

— Reply to this email directly, view it on GitHub https://github.com/immunogenomics/harmony/issues/240#issuecomment-1945641505, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSFW2H54VBDHQ6BYPJBCOTYTXFB5AVCNFSM6AAAAABCXNJDCSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNBVGY2DCNJQGU . You are receiving this because you commented.Message ID: @.***>

[DarioS]

I am unsure if cancer cells from tissue and cell culture should merge. That is a tough biological question. I have not seen anyone else with the same kind of paired data. Is there such a data set within Broad Institute which you could try? I can't share the one which I am working on. Cell culuture contains only two cell types; cancer and cancer-associated fibroblast. Tissue contains eight cell types. Wouldn't lower values of theta be recommended? I want the data to merge more, not less. I tried 0.

[pati-ni]

Can you share UMAPs before and after the integration of the two datasets? Also, theta needs to be increased to force the cells together. Setting it to zero disables the low batch diversity penalty.

[DarioS]

Ah, O.K. increasing theta is the solution which leads to good mixing.