In　order　to　compare　orthogonal　encoding,　five　byte　encoding,　Codon　encoding　(two)　and　Profile　encoding　to　find　out　their　virtues　and　shortcomings,　we　carefully　studied　them　with　artificial　neural　network.　Results　indicated　that　Profile　encoding　that　preserves　the　redundant　evolutionary　information　got　higher　prediction　performance,　and　that　how　to　use　biologic　evolutionary　information　effectively　is　very　important　to　improve　accuracy　of　protein　secondary　structure　prediction.