re-producing the figures in MAGIC paper

[krishnaswamylab]

re-producing the figures in MAGIC paper

[dvdijk]

Hi,

We only used the 'unsorted' data. In addition we used the data that the authors of the original paper had used for clustering. These clusters were not published so they sent them to us. However, I'm not sure I can share these.

MAGIC settings:
npca = 100
k = 12
ka = 4
epsilon = 1
t = 9

David

[radio1988]

I see. Thanks for the quick reply. What do you mean by 'unsorted' data? And, maybe the authors of the original paper filtered out cells with less than 1000 reads?

[dvdijk]

https://www.dropbox.com/s/v0v7fhf8foryacz/MAP.csv?dl=0

[radio1988]

Thank you David! So selecting these 2730 cells from 'GSE72857_umitab.txt.gz ', and filter out genes with no reads among these 2730 cells, then use the magic parameters you just showed us and we should re-produce the same result?

[dvdijk]

right

[radio1988]

thanks!

[radio1988]

Hi David,

We selected 2730 cells from the raw matrix, normalized the data, did magic, and plotted gene-gene relationship. The result is very similar to the figures in your paper. Thank you for your help!

I found that in order to get similar result as in your paper, we don't do log-transformation. Is that what you suggest?

Thank you!

[dvdijk]

that's correct. we only did library size normalization

[radio1988]

Thank you, the re-producing those biaxial plots were successful!