I am studying a nuclear protein and want to come up with a list of potential proteins that it interacts with. From the nuclear fraction of 293T cells, I did an IP (immunoprecipitation) to pull down my protein and submitted whatever I pulled down for a mass spec analysis. I also did an IP with a non-specific antibody as a control. I got a huge list of hits, and I'm not sure how to prioritize them. I can manually remove the hits that also came down with the non-specific antibody, those are obviously just sticky proteins. I looked around Pubmed for ways to prioritize non-quantitative mass spec hits, but I'm not sure if any of them would work in my case. All of them require multiple mass spec repeats - I can do that if I need to, but would like a way to look at my initial data more intelligently.
If they are available, talk to the people who analysed your data. Using their knowledge is intelligent.
I will assume two critical steps are completed. First that your results have been searched correctly. I will also assume that the peptide and protein identifications have been filtered to provide a false discovery rate of 1:100 at each level.
You can remove all the proteins that were identified in your affinity purification control. Another option is to use the Contaminant Repository for Affinity Purification (the CRAPome) which provides a formal framework for removing contaminant proteins. It is a nice resource if your data are compatible.
The remaining proteins are what you have to work with, this is the
list of potential proteins that [your protein] interacts with.
Prioritise these proteins based on the biology of your experimental system. If the list is small work through the function of each protein and assess how it fits with the biology you are researching.
To assess the potential functional links between the proteins you can create an in silico network. Markov clustering can highlight the presence of functionally distinct groups of proteins within a network. String, linked above, can do this. Centrality measures can indicate the presence of proteins with the potential to influence the behaviour of the network.
For larger lists you can also assess the function of the proteins in the list by performing a functional annotation enrichment analysis. There are many other things you could do but the primary goal is to keep biology at the forefront of your efforts.
Finally you mention that repeats are necessary. Analysing the data is time consuming. If this was a test run verify that the test was successful and then run the replicates. Time spent analysing incomplete data can be spent completing the data set or ensuring the right data is collected. Quantitative data can be very useful and with care shotgun proteomics approaches can provide quantitative data but the people who run the machine need to be consulted when quantitative data is required.
You can look for the most abundant proteins in your sample. Although, as you say, the LC-MS analysis of your IP was non-quantitative, you can still try to get some quantitative estimation. To elaborate, you can use for example sequence coverage, no. of PSMs for each protein, no. of peptides ID'ed for a protein, share of spectrum ID, etc, as proxies for abundance.
There exists rough correlation between abundance of a protein and all these quantities. If a protein is more abundant you are more likely to have more number of peptides ID'ed from it, more no. of PSMs assigned to it, it would have higher sequence coverage, etc. Different people use different things. I have often used no. of PSMs normalised by length of the protein as an estimation for the abundance of that protein.
As others mentioned, discuss with the people who analysed your data. As a rule of thumb, I would filter my data first on 1% FDR at the PSM level, then at the Peptide level and then at the protein level. Then I would throw out all the proteins that were not identified by at least two unique peptides. Then I would compare the proteins ID'ed in the negative control list and remove the common IDs. Then sort the remaining list on no. of PSM for each protein. This would give an estimate of the more abundant proteins in the IP sample. To give an example, your bait protein (protein against which you did the IP) should be the most abundant protein in your sample. Proteins after this could then be assumed to be interactors of your bait.
Do remember that this is not a proper quantitative way of measuring protein-protein interactions by affinity purification coupled to LC-MS. But, it is a useful start. I always used preliminary information from such an analysis to design better ones. Also, read about reciprocal IPs to confirm interactions.