<"...an algorithm that could predict physical characteristics based on DNA, and even more importantly perhaps, to predict DNA based on physical characteristics."

Think about that for a minute or two...

Fossils are stones, not bones. There's no DNA to be recovered, much less sequenced, analyzed and plugged into some fancy algorithm. DNA is also notably unstable; preservation over millions of years takes extremely special conditions and plain dumb luck. Even then, you get fragments a few dozen or hundred base-pairs long, not entire genomes. The difference between the human and chimpanzee genome sequences is widely quoted to be approximately 2%.

Working in the other direction... you are basically asking a software engineer to write code to attempt what palaeoanthropologists already do: assign genus, species, and historical relationship to a handful of former bone fragments, based on inferred characteristics of the total animal. Where's the code geek going to get the expertise for his "expert system"? The experts, of course... who already can't determine whether this new find is or isn't a hominid, or if it is, whether or not it's in the "direct line" of human evolution.

Work on a time machine instead... Even then, don't count on DNA sample returns, or even live specimens, to answer the question, "Is it Gramma?"

Fox