Interphase 2: Non-quantitative peptides are a feature, not a bug.
As discussed in the previous post (Interphase 1), silica enrichment removes most free protein from XRNAX extracts so that quantifying their protein-content using SILAC can be used to quantify RNA-binding between conditions. In our manuscript we validate these quantitative features of XRNAX in the experiment presented in Figure S11G.
As you can see in the middle graph, although most peptides are accurately quantified, a fair number of peptides do not show any change but are equally abundant independent of crosslinking. This result could have been predicted from Figure 1F, where some peptides were very strongly enriched after silica enrichment, whereas others were not.
For a setup that compares two UV-crosslinked samples without non-crosslinked control, this should mean that previously super-enriched peptides quantify differences in RNA-binding adequately and previously weakly enriched peptides don’t. As you can see in the right graph of Figure S11G this is the case. Interestingly, non-quantitative peptides do not randomly distribute but form a pretty narrow distribution around 0. Consequently, non-quantitative peptides in an XRNAX experiment can be used to correct for mixing errors between the SILAC channels. Let’s consider an XRNAX experiment where 10 million SILAC heavy cells were treated with arsenite and 20 million SILAC light cells were left untreated as a control. Both populations are UV-crosslinked, XRNAX extracted, silica enriched and quantified by MS. Non-quantitative peptides will differ by a factor of 2 between the SILAC channels, so that you can correct for that foldchange, even if you did not know that different amounts of cells had been used. In praxis you can usually use the normalized peptide ratios returned by MaxQuant because MaxQuant will have corrected these ratios according to ratios of the most dominant population in the measurement. However, you cannot use these corrected values if non-quantitative peptides are in the minority. That means for experiments where you expect most proteins to change their interaction with RNA you should make sure that MaxQuant, or whatever software you prefer for quantification, used the right population for the normalization. In any case you will need to discard ratios for non-quantitative peptides and only keep ratios of quantitative peptides, which you previously identified in an experiment using a non-crosslinked control. Note that a list of super-enriched peptides compiled from the cell lines MCF7, HeLa and HEK293 can be downloaded from the Applications section of this website. However, we cannot guarantee that in your particular setup these peptides will behave quantitatively, too. Therefore, we recommend you record your own non-crosslinked control dataset in order to find out which peptides you are able to super-enrich. We do not recommend using an experimental setup with three SILAC channels, which includes a non-crosslinked control apart from two UV-crosslinked samples, because without the MaxQuant ‘requantify’ feature that will give you very few ratios. Using the ‘requantify’ feature, knowing that one SILAC channel is supposed to be empty, doesn’t make any sense, for why we recommend recording your background separately.