Contacting a searching company is probably the most efficient way to do a comprehensive search. I have a contact at Thompson Reuters if you are interested.

Dear Brian,

thank you very much for your kind offer. Outsourcing is certainly a potential solution. But my question was not so much about who could do it for me but how you folks would approach this.

I am assuming this is a search for purposes of patentability only. Do the ~500 possible compounds have a common core structure? If so, I might start with a substructure search on the CAS Registry database. That should identify any compounds known to exist which have that core structure.

A manual search of the patent art is also useful. Chemical compound patents are classified based on structure, so that a search of the relevant subclasses is often feasible. In the case of ~500 compounds, it may be difficult unless the compounds are very similar. If not, we might request that the inventor identify a smaller number of compounds to serve as the basis of the manual search. We also do keyword searching on databases such as Derwent and Espacenet.

Another possible issue might be hypothetical compounds encompassed by Markush groups described in published patents. A search for Markush structures in the prior art can also be carried out on the MARPAT database. We also review Markush structures uncovered during a manual search to see if they cover the compounds at issue.

The searching starts with Exact structure search in Chemical Abstract registry. Then I will expand my search to similarity and substructure searches for the whole structure or core of chemical structure. For Markush groups the we can search on MARPAT database/MARPAT file. We can also search in Derwent, Thomson, Patbase for key word search combining with specific IPC codes. But most useful prior art searching of chemical structure is in CAPlus/ CAS registry.

We rarely expend time or money on prior art searching. A researcher engaged in a specific area should be aware of much if not all of the prior art. If you think it is novel and useful, it may well be. The cost of filing a patent application is $10,000 - $15,000 ± a lot. With that given, it is not reasonable to spend several thousand dollars (a practical minimum) on a search. And were one concerned, the searching should be done before embarking on research.

If you have to ask, then your claim will fail, for lack of novelty/inventiveness, or for being designable-around. Tell your client to bring in someone (not me) who knows what he is doing. J

Some firms as you've just seen don't search knowing that filing less than stellar applications will require a lot of back and forth with the patent office, for which you must pay. It shouldn't be a crap shoot: know for sure nobody has done it, and address anything close in your application. Always search thoroughly from the outset: it will let you write the best specification and write around the art the examiner will find anyway (you're stuck with that first specification, you know-- you can't improve it later).

Searching helps you focus on the technical language and gives ideas on what to include to nail down the case. If you do it yourself, don't stop until you find something uncomfortably close-- then search for more like that one, from the same inventor or in the same class. Attorneys are rarely the best at the art (they often outsource it too): just type "patent searcher" in the search box, and welcome to a whole new world of specialists.