Logfold change

Question

[kumar.ajneesh]

I have three categories of data and each category has two samples. How can I calculate the log fold change for this data?

Thanks in advance for your help

Kumar Ajneesh

[ATpoint]

What "data"? It matters, and at best there is a tool that does elaborate stats under the hood while for the end user it is running some one-liners.

Answer 1

[jared.andrews07]

mean(log2(group1 samples)) - mean(log2(group2 samples))

Repeat as needed for whatever comparisons. As noted by rpolicastro, add pseudocounts to each value if any of the means are zero.

[rpolicastro]

If any of the means are 0 you should add a pseudocount to all values too.

[Gordon Smyth]

That is not the log fold calculation used by any popular software tools that I know of. For microarrays, PCR or mass spec, replicate expression values would be logged before averaging. For sequencing data, the log fold change calculation would be more complicated and would need to take account of library sizes and count sizes.

[jared.andrews07]

Good points. I've edited my answer to address the first.