Protein inference in peptide-centric mass spec proteomics? Quantitative proteomics? PTM analysis?
Protein inference from peptides. It's an old problem that still is not solved (and don't think won't be in a long time). There is no one solution but there should be more choices at how to look at the peptides in order to mine more meaningful information.
Missing spectra. In an average experiment LC/MS/MS at most 30% of MS/MS spectra. The current focus is on getting rid of false positives but there is still a lot of good quality data that can't be explained. This is specially important with the new generation mass spec instruments that achieve a very good scan rate (~50 scans per second) with very high resolution. I think there'll be a trend to get back to those good quality spectra to see what they are.
High resolution of current mass spectrometers is still not fully exploited. I foresee new methods to improve mass accuracy with recalibration.
Making proteomics mass spec data more accesible for biologists, not only proteomics people. After all I see mass spec proteomics is a technology to get more insights in biology. Many high-throughput studies with spurious identifications have made the field to lose some credibility. The proteomics field should gain the respect of biologists, not only the proteomics community. As a proteomics informatician I feel responsible of making software more useful for biologists, not only proteomics people.
MRM is going to be a big thing. For that the databases of transitions need to be in place.
Labeled quantitative proteomics losing momentum. The correlation with transcriptomics (relative, not absolute) data is good enough. Unless you have a very specific experiment, in general, if you compare the cost and the time of a transcriptomics experiment with a quantitavive proteomics experiment, the transcriptomics experiment is a more sensible choice for a biologist. At the same time I see label-free getting easier, but for that the sofware has to improve.
Targeted proteomics experiments instead of global ones. Instead of comparing 2 cell types in 2 states and try to find what's going on, I see more hypothesis driven experiment coming back to proteomics. For example, protein interactions with co-IPs, looking for specific PTMs in a set of known proteins, organelle proteomics, etc... It's hard with current proteomics software to look at the data with a targeted approach.
Proteogenomics: using mass spec data to annotate genomes. After all with mass spec data you can get direct evidence of protein translation.
De Novo proteomics. Current identification methods rely too much on cDNA databases. I bet there is a hidden proteome out there that can't be directly inferred from the protein databases. Ligases, programmed frame-shifts, low abundance stable mRNAs, etc...
Alternative fragmentation methods. Currently CID is the king, but HCD and, specially, ETD are complementary fragmentation techniques which are not fully exploited because of the lack of software.
Phosphoproteomics is hot right now but I see other PTMs are going to get as hot as phospho is now.