Over on io9, there's a nifty article by titled The Great Lost Manuscripts of Science Fiction and Fantasy. It leaves out the original Riverworld manuscript, which Philip Jose Farmer later claimed had an eminently satisfying ending which he'd completely forgotten by the time he rewrote it and penned the subsequent sequels. But otherwise, it hits the high points.
Alas, most of the lost manuscripts deserved to be lost. Given how robust the SF short fiction market of the 1930s and 1940s was, any story that Isaac Asimov couldn't sell had to be pretty dire. And Robert A. Heinlein spent decades tracking down and destroying every copy of his first novel, We The Living, he could find. (When it was posthumously published, it turned out to contain most of the ideas that later made his reputation in a tedious and didactic plot.)
When Jules Verne's long-lost novel, Paris in the Twentieth Century, which was turned down by his publisher as being far too unlikely, was finally published however, it turned out to contain an astonishing number of perfectly accurate predictions. His record there may have been better than that of H. G. Wells. Which would have pleased Verne mightly.
The year the book was published, my friend Allen Steele and I independently recommended it for the Nebula Award. Making us the only two human beings in history ever to vote for Verne for a Nebula.
We have mingled blood, Allen, and shall always be brothers.
You can read the article here.