There are 120 data files in this archive (plus this README).
The data consist of true acceptor sites (celegans_acc*+1.{train,val,test})
true donor sites (celegans_don*+1.{train,val,test}) and the
decoys (celegans_{acc,don}*-1.{train,val,test})
The training was done on the *.train only files. The *.val files were
used to adjust the model parameters (like C for the SVM and the
number of states for the HMM) and the testing was done on the *.test
files.
Training was done on 100, 1000, 10000 sequences (*tr100*, ..).
The validation data consists of 5000 samples, whereas the
test data has 10000 samples.

All results are averaged over 5 runs (*run1*, ...).

NOTE: We removed all sequences containing characters other than ACGT.
