Hi everyone,

I am currently working with peptidomics MS data from patients with and without disease, and I would appreciate some advice regarding quality control.

My understanding is that, when the data are initially generated from MS, many values are actually missing values (NA), but in the matrices I received these missing values were replaced by zeros.

I believe I should perform QC both at the sample level and at the peptide level. My initial matrix contains around 5,000 peptides, and there are quite a lot of samples with a very high number of zeros and relatively low total intensity. For example, some samples have more than 90% zeros and only a few hundred detected peptides.

My main questions are:

1. Is there any commonly used sample-level filtering rule in peptidomics for removing poor-quality samples? For example, removing samples with more than 90% zeros, very low numbers of detected peptides, or low total intensity?

2. Would it make more sense to define sample QC thresholds globally across all samples, or separately within each biological group? I also tried IQR-based rules, but I am unsure whether QC should be done on all samples together or stratified by group.

3. PCA has not been very informative in helping me decide which samples to keep. Is that common in this type of data, and are there other QC approaches that are usually more useful?

At the peptide level, I already removed peptides that are zero in all samples, but there are still many peptides detected in only a small fraction of samples. I decided to keep only peptides detected in more than 60% of samples in at least one group. Does this sound reasonable, or would you recommend a different filtering strategy?

Any suggestions, references, or examples of common QC practices in peptidomics would be very helpful.

Thank you very much.