This can be a very challenging task, but there are some steps you can take in the right direction (one of which you already have with checking for 'WH' words at the beginning). It also heavily depends on the domain of your 'sentences' (search queries, utterances, etc.).
One thing you might try is to (1) do a thorough check for obvious questions as you already described. This would include checking if the first word is either a WH word OR an auxiliary verb ('is', 'should', 'will', 'could', 'can', etc), since in English often times questions undergo subject-auxiliary inversion. (2) If you have any natural language processing libraries available, use POS chunking and look for sentences that consist only of a Noun Phrase (NP). These, since they contain no predicating information, are generally inquiries about the Noun Phrase. I'm not sure what you are using for your analysis, but one great POS chunk analyzer is that of the Pattern nlp module for python (http://www.clips.ua.ac.be/pages/pattern-en#parser). Hope this helps.