I'm working on a university project of predicting Translation Initiation Sites in human DNA. I searched the net for papers and documentation to get guidelines and inspiration, but I feel uncertain that I was able to find the state of the art in solving this problem, so I thought of asking the community here: What's the state of the art method of predicting Translation Initiation Sites?
(I'm also posting the question to help this nascent Q&A site by participating, please close it if it completely misses the mark of being a good question.)
Ribosome Profiling experiments profile the positions of ribosome-protected mRNA. a cycloheximide treatment fixes the translating ribosomes. A naive approach to define Translation Initiation Sites (TIS) would be to look at positions where you see distinct 'peaks' in the distribution of these fragments since the ribosome density will be higher at the start of the CDS (at the TIS).
TISDB is a database comprised of GTI-Seq global mapping of TIS codons at nearly single-nucleotide resolution. GTI-Seq uses two 'fixers' that fix ribosomes where one of them specifically fixes the initiating ribosomes.