infercnv heatmap appearance of 'Reference' (non-malignant) cells: opinion required

[jjacob12]

Hi, we performed scRNA-seq on human iPSC-derived brain organoids in co-culture with tumour cell-line derived malignant brain cells. The latter cells were not clonal. I performed infercnv analysis using the standard workflow and obtained the attached heatmap

There are equal numbers of cells in the 'References' (non-malignant) as there are in 'Observations' (malignant cells). There is a very obvious difference between the infercnv heatmap profiles, as expected. There are also cells in the 'References' with a weak infercnv signal. I have two alternative explanations for this: 1. could be noise(?) because we recently tested over 600,000 genomic sites of the iPSC line using a SNP array and could not detect any gross chromosomal amplifications, deletions, or re-arrangements. Moreover, iPSC-derived organoids were maintained in culture for only 60 days and were largely maintained in differentiation medium over that time. 2. Organoids contained multiple distinct cell types undergoing differentiation and differentiated cells. The expectation would be that the latter cells would express or repress specific sets of genes that define their identity, so I wonder if it is perhaps not that surprising to see some 'weak' CNV signals in the 'References' cells?

[alexwskh]

Hi, not part of the team or a bioinformatics expert but thought I would chime in. Could you explain a little bit more about how you collected the cells before sequencing? Did you collect them all together? Since you defined the reference cells in this set, how did you do so? Were some marked with an obvious reporter? If you aren't 100% confident about your reference cells, I think the easiest explanation is that you have some malignant cells in your reference batch.

If you are feeling confident about your reference cell selection, I think your explanation makes sense. It feels intuitive that a family of genes involved in a certain transcriptional program might be located near each other on the chromosome and regulated together. There is some evidence for that as well:

https://pmc.ncbi.nlm.nih.gov/articles/PMC4150778/
https://pubmed.ncbi.nlm.nih.gov/28835372/

However it might not be an "always" thing:
https://pmc.ncbi.nlm.nih.gov/articles/PMC8457456/

[jjacob12]

Hi there and thanks for getting back to me. The 'Reference'-only non-malignant cells (as one control) were grown in completely separate culture dishes from the malignant ones. Additionally, we also had Reference cells in co-culture with malignant cells. In the figure I attached above, the Reference-only controls could not have been mixed up with the malignant cells as the two cell types never came into contact. Reference-only controls and cells grown in other conditions were processed for scRNA-seq
on the 10X platform in the same run and then the same flow cell of an Illumina sequencer. So, yes I'm confident about the Reference cell selection and it's very interesting there's some evidence to back up my speculation as to the weak infercnv signals in this batch of cells.