jordangumm / nucompress Goto Github PK

Results in less than ideal compression for short sequences, even though they appear to be 1/4 the length when printed out.

e.g.,

>>> from nucompress.compression import TwoBitCompression
>>> compressor = TwoBitCompression()
>>> seq = 'CTGACTGGCTGACTGG'
>>> encoded_seq = compressor.encode_nuc(seq)
>>> encoded_seq
{'length': 16, 'seq': 'xzxz'}
>>> type(encoded_seq['seq'])
<type 'str'>
>>> from sys import getsizeof
>>> getsizeof(encoded_seq['seq'])
41
>>> getsizeof(seq)
53

This is because Python strings are 37 bytes + 1 byte per additional character. Source: https://stackoverflow.com/a/30316760/31248

At the moment 2-bit encoding ignores sequences with N values. According to the Algorithms for Next-Generation Sequencing Data chapter on De Novo NGS Data Compression, the N can just be ignored. https://link.springer.com/chapter/10.1007/978-3-319-59826-0_4

The N symbol (that actually rarely appears) does not necessarily need to be encoded because its quality score is always 0. During decompression, the symbol is simply inserted when a null quality score is seen. Packing four nucleotides per byte represents, however, a good compression rate, decreasing the size of the DNA sequences by a factor of 4.