Hi,

I am interesting in clustering my single cell data to identify clusters and then figure out if certain genes differ in expression profiles and then try to link them to developmental stages. 

1. So, to be able to compare gene to gene expression, I suppose the expression scores should be corrected for gene length. Does SC3 do this somehow? I don't pass in a gene length argument at any point, so I was wondering about this.

2. I am using salmon count data (using tximport) for SC3. Now I was wondering if I could use salmon 'length scaled tpm' for SC3 clustering. Does it make sense to use this? If yes, how would I go about doing that?

Thanks, Roy

Hi,

1. SC3 does not correct for gene length, please use other tools for that.

2. Yes, you can use any units with SC3, it will always find some structure in your data. The thing is that if you don't normalise your data, or normalise it in a wrong way you will either get a wrong result or pick up a batch effect. So please make sure your data is properly normalised before running SC3. If you trust salmon 'length scaled tpm' then go with it.

Cheers,

Vladimir