By valid binding proteins, do you mean real/true interactors ? There's no sure way of inferring protein truly binding to the bait (directly or indirectly, vs random interactors/contaminants) from a single IP. Every protein identified as being pulled down with the bait is a potential true interactor. People have used various heuristics to try and increase confidence but this often requires more data than just one IP. You could also clean up your list of potential interactors by removing known "contaminants", i.e. proteins known to be non-specific interactors. Such lists are available for example from the crapome database or you could build your own based on your lab's experience or by doing control IPs. With more data such as replicates or reciprocal IPs, you have more possibilities for identifying contaminants. How to proceed would depend on what data you have.