The new issue of Nature Methods contains an article outlining a European consortium which is resolving to generate a vast array of affinity reagents targeting the human proteome. It is a daunting task, and most of the article is tightly packed enumeration of what is daunting about it.
Even deciding what to generate reagents to is no easy task: take the 25K human ORFeome and mix in various post-translational modifications (proteolysis, phosphorylation, glycosylation, ubiquitination (and its cousins SUMO, NEDD8, ISG15 and several more), acetylation, lipidation) and variant folding/activity states, and there are almost certainly more than the 100K targets that they propose going after -- which is one of the first statements in the paper.
Their review of available technologies is brief, but gives a quick listing of about all the choices -- none of which has been proven to be scalable to the task. There's even the question of whether heterodox scaffolds, such as non-antibody protein scaffolds or nucleic acid aptamers, are appropriate or whether antibodies are the only way to go. Deciding on what assays to support for validation is no easier than deciding what to target, and doing validation on that scale may be a bigger challenge than generating all the reagents in the first place.
They also propose wrangling the available information on binders, in particular keeping track of where things bind. The current state of affairs is terrible, with many antibodies labeled only with common names by the ever expanding multitude of antibody suppliers. Worse yet, the same antibodies are sold by multiple vendors, but with no consistent way to tell -- except by looking for identical background noise in the Western blot images from each vendor. Sites such as ExactAntigen try to get a handle on things & cut down the tedium of antibody identification a bit, but the antibody information world is more chaos than order.
All that said, boy is this project needed! They will need a lot of luck (and copious quantities of money), but the lack of off-the-shelf affinity reagents for any protein of sudden interest is a serious handicap for validating array or computational experiments, as I found all too often in my last job. DNA & RNA have the elegant beauty of Watson-Crick pairing; if only such a rule system could be devised for protein!
BTW, if you don't get Nature Methods, the print subscriptions are given out pretty freely to biotech/biopharma professionals.