Detailed description of the FASTA format is available at NCBI.
Important notes:
Empty lines are NOT accepted.
At present, the sequence input limit for first two (faster) classifiers is 2000 sequences.
Due to computational intense of PSI-BLAST searches against the Swissprot or Uniref90 databases and their corresponding conversion to PSSM matrices, the current limit for PSSM based and AAindex + PSSM based classifier are 1000 or 280 sequences per submission, respectively. For larger analysis, please contact us.
Each sequence should be at least more than 20 residues in length. The shorter sequence might be a "fragment" and so, the prediction will not make any sense.