Tweet
Hey,
I've been working with some mouse faecal samples provided to me and was happy enough that they had been classified well by Centrifuge. Then, when I came to actually start using the data I realised I had thousands of classifications which were giving read and unique read counts from single digits to many thousand but an abundance of 0.
I don't know the Expectation-Maximisation algorithm which Centrifuge uses to calculate abundance well enough to know if this is reasonable. To my novice mind it shouldn't matter what model you use if the tool detects the presence of the OTU classified the abundance should never be just 0. Infinitesimally small I could understand, but not a flat nothing.
Any advice on the subject would be appreciated.
I've been working with some mouse faecal samples provided to me and was happy enough that they had been classified well by Centrifuge. Then, when I came to actually start using the data I realised I had thousands of classifications which were giving read and unique read counts from single digits to many thousand but an abundance of 0.
I don't know the Expectation-Maximisation algorithm which Centrifuge uses to calculate abundance well enough to know if this is reasonable. To my novice mind it shouldn't matter what model you use if the tool detects the presence of the OTU classified the abundance should never be just 0. Infinitesimally small I could understand, but not a flat nothing.
Any advice on the subject would be appreciated.