The order of source sample could influence the final output ?

cozygene / FEAST

Fast expectation maximization for microbial source tracking

Other

115 stars 60 forks source link

The order of source sample could influence the final output ? #6

Closed JX-Zhong closed 5 years ago

JX-Zhong commented 5 years ago

I have 2 source samples, and when i change the order of these two samples, i found that the output changed a lot. Why? Selection_057

liashenhav commented 5 years ago

Hi,

Can you please share the input files? my email is liashenhav@gmail.com.

Thanks,

Liat

liashenhav commented 5 years ago

Hi,

I was able to replicate the problem. It seems that the sequencing depth in some of your sink samples is very low in comparison to the sources (e.g., sink 1 = 2,124, NC3_RNA = 1,262,457, NC1_RNA = 924,525). As a result, the noise added to the model when initializing the unknown source is too extensive. This is now scaled, such that it takes under consideration the sequencing depth as well as the similarity between sources.

Thank you for bringing this issue to our attention.

Liat