fastTopics on normalized data

[nesilin]

Hi!

Thank you very much for developing such a great package!!

I'm planning to use fastTopics to extract common gene modules across different tumor samples. I'm working with Visium data in a very heterogeneous cancer. Since I have variability between and within samples regarding UMI counts and cell density I figured library-size normalization of the the data might help overcome this. However, in your paper and the vignettes you explicitly insist on the usage of the raw counts. What are the risks of using log normalized values?

Thanks, very cool tool ;)

[pcarbo]

@nesilin Thanks for the positive feedback.

The topic model is based on a multinomial model of the UMI counts which conditions on the total count for that cell, which is related to sequencing depth. The Townes et al paper has a more detailed discussion of the use of the multinomial model for UMI counts. So in principle this should handle variability in sequencing depth. However, I should point out that the topic model may not be able able to deal with the inter-tumor heterogeneity in your cancer data. For this we are developing new methods to better identify gene modules from multi-tumor cancer data (I'm asking my colleagues now and will share a link when I have it).

Hope that helps.

[nesilin]

@pcarbo thanks for the quick answer. We will keep an eye on the mentioned upcoming tools!

[pcarbo]

@nesilin Apologies for the late reply, but we have finally posted the new methods here, which are also described in this paper.