LEfSe output file

[Liam Shaw]

Hi,

I have a question about the output of LEfSe. 

The output of running the LDA effect size calculation is given in columns. I assume they are categorised as follows: 

feature | ??? | class | LDA effect size | p-value

What does the second column represent?

Thanks,

Liam

[Nicola Segata]

Hi Liam,

 the second value is the log of the highest class average.

I hope this helps

thanks

Nicola

[bala krish]

Dear Nicola,

Does the high LDA score valid if it is close to the log of the highest class average??? Or it does not matter.

f_961454	2.49533401767	XXXXXXXXXX	2.34131978395	0.00534567687265
f_1088305	3.72917146705	XXXXXXXXXX	3.08799347456	0.0465329850745
f_988314	3.45609455262	XXXXXXXXXX	3.12131905416	0.0406089734043
New_ReferenceOTU102198	2.04054680989	XXXXXXXXXX	2.30237992075	0.0342660560639
f_711184	2.06933058862	XXXXXXXXXX	2.95350821186	0.00534567687265
New_ReferenceOTU156461	1.41260313717	XXXXXXXXXX	2.74842348347	0.0186029299011

Bala 

[Nicola Segata]

Yes, it is valid. The LDA score is a measure of effect size which has some relation with the differences between the class averages.

cheers

Nicola

[jian wang]

Hi Nicola,

I am new to LEfSe and having questions regarding to the output file for the LDA effect size. 

For the second column, it is the log of the highest class average. Since the data is relative abundance (<1), when I tried to calculate the log of the class average, I always obtained a negative value. I am wondering why all the values at second column are positive. 

For the last column, which test is the p-value corresponding to? The KW sum-rank test? The pairwise Wilconxon test?

Thank you very much!

Jian