Hi Ase,

I think log2 fold change (FC) is quite convenient for doing plots, since it distributes the differences equally around 0. While for non-log FC (ratio) for A/B, if A is greater than B, it will be >1 but for B greater than A, it will be a value squeezed in between 1 and 0, right? Therefore it is hard to tell what the ratio is unless you do the reverse opperation for those (B/A). Also, many of us is already used to log2 FC differences.

I would say, if you are mostly interested in visual representations of your data between both groups, you can try volcano plots, which very intuitively show you the most significant up/down-regulated hits (y-axis: -log10 p-value; x-axis: log2 FC). However, these plots remove the information on which are the most abundant proteins, since it could be that a protein with FC=5 and p-value=0.0005 is just in the mid- or low-range of abundance. You should also consider the abundance of the selected significantly changing hits within the sample of interest, and ponder whether a higher- or lower-abundant protein with signigicant changes may be more interesting/useful for clinical purposes.

Hope this helped.

Cheers,

Julia

Thank you, Julia! This was truly helpful.

Best wishes, Åse

Dear Ase,

If I'm not mistaken the Fold change is calculated from log2 values by subtraction. Due to the logarithmic calculation properties log2 m - log2 n = log2 (m/n) [where m and n refer to your samples] , in the end you do get a ratio of the two absolute values, but it is logarithmized which is more suitable for presentation as Julia described before.

If your biomarker data is already calculated from other raw data before the log 2 transformation, i think you can still use the fold change, but you should pay attention to draw the right conclusion.

Hope i could help.

Good luck with your project.

Peter

Any log will work. Most people think in Log10 (common log) because it implies order-of-magnitude (factors of 10) differences. Log2 implies doubling which is also easy to grasp. Presenting data on a log2 basis, however, magnifies what may be minor sampling variations and the number of prospective differences. Presenting data on a log10 basis will point out only the order-of-magnitude differences between samples and result in fewer prospective differences. Since the name of the game is to find something that others missed so that you can get a publication, log2 is often favored, even though more of what is found may be false positives.