To avoid error messages when converting the original FASTA file into database binary files, you should make sure that all sequence lines are of the same length except for the last line of each sequence which could be ragged (or sequences can be stored on a single line with no breaks whatsoever).

If the original FASTA text file with nucleotide sequences has name NAME, program PRESSDB produces the following 3 binary files:

• NAME.nhd -- all sequence headers written one after another without any special encoding or compression, each header starts with symbol '>', there is NO other '>' within one header, so that, if necessary, the sequences can be counted by the number of  '>' sumbols in the header file.

• NAME.csq -- nucleotide sequences written one after another in a compressed form, sequences are separated with a "magic byte" which has value 0x78, that is '-'; four nucleotides are stored in one byte, that is two bits are used per one nucleotide; the correspondence between nucleotides and their binary codes is as follows.

 

Nucleotide	Encoded as	Nucleotide	Encoded as
'A'	0 (binary 00)	'W' ('A' | 'T')	8
'C'	1 (binary 01)	'S' ('C' | 'G')	9
'G'	2 (binary 10)	'B' ('T'|'G'|'C')	10
'T'|'U'	3 (binary 11)	'D' ('A'|'T'|'G')	11
'R' ('A'|'G')	4	'H' ('A'|'T'|'C')	12
'Y' ('C'|'T')	5	'V' ('A'|'G'|'C'))	13
'M' ('A'|'C')	6	'N' ('A'|'T'|'G'|'C')	14
'K' ('G'|'T')	7	'-'	15

Note, that unlike similar amino acid file, NAME.csq contains sequences that are on average longer than the true sequences in the original FASTA file.  This is because when sequence length is not divisible by 4, there are inevitably a few unoccupied bits in the last byte of the compressed sequence, and there is no way to distinguish these bits from encoded 'A' nucleotides.

Important: degenerate symbols (such as 'N' -- any nucleotide, 'R' -- a purine, or 'W' -- a weak-bonding nucleotide, etc.) in the compressed nucleotide sequences are randomly substituted with a non-degenerate symbols from the corresponding set of nucleotides; the only way to get the original nucleotide sequence is to go to the original FASTA file.)
Finally,
 
• NAME.ntb (table with absolute file addresses of compressed sequences in file NAME.csq, and sequence headers in file NAME.nhd).

• 32-bit int -- database type.  For a nucleotide database this number should be set to 0x788325f8.
• 32-bit int -- database format.  For a nucleotide database this should be set to 6.
• 32-bit int -- the number of bytes in the title of the database (N).
• N * (8-bit byte) -- database title.
• if (N % 4) not equal to 0, skip (4 - N % 4) bytes.
• 32-bit int -- the length of line in the original FASTA file (Important: all sequences in the original FASTA file MUST be formatted in the same way, otherwise PRESSDB wouldn't work).
• 32-bit int -- the number of sequences in database (M).
• 32-bit int -- the length of the longest sequence in the database.
• 32-bit int -- the number of characters in the database BEFORE sequences are COMPRESSED.
• 32-bit int -- the number of characters in the COMPRESSED database.
• 32-bit int -- the number of "overrepresented" nucleotide 8-mers (N8). If this number is non-zero, corresponding 8-mers will not be used by BLAST for searches in this database.
• skip (N8 * 4) bytes (or read overrepresented 8-mers if required).
• for(i=1; i less or equal to M; i++) {
read 32-bit int
(= the absolute position of the beginning
of the ith compressed sequence in file NAME.csq)
};
• for(i=1; i less or equal to M; i++) {
read 32-bit int
(= the absolute position of the beginning
of the ith sequence in the original FASTA file)
};
• for(i=1; i less or equal to M; i++) {
read 32-bit int
(= the absolute position of the beginning
of the ith sequence header in file NAME.nhd)
};
• Byte array of size (M/8 + 1), where ith bit is 1 if the ith sequence has a degenerate symbol, and 0 otherwise.
• end of file.