Skip to content

DNAFingerprint

DNAKit currently offers two methods of DNA fingerprint calculation. Each bit has only 0 or 1, which can be directly used for Tanimoto/Jaccard similarity, retrieval, clustering and deduplication.

1) Hashed k-mer bit fingerprint

  • Purpose: Extracts k-mers occurring in a sequence and stably hashes them into fixed-length 0/1 bit vectors for use in Tanimoto/Jaccard similarity, clustering and retrieval, while avoiding excessively large explicit 4^k feature spaces.
  • Method: Hash k-mers into n_bits positions and set observed positions to 1; collisions can map distinct k-mers to one bit.
  • API: dnakit.fingerprints.hashed_kmer_fingerprint(value[required], k[required], n_bits[optional], canonical[optional], seed[optional], representation[optional], ambiguity_policy[optional], overlapping[optional], cross_gaps[optional])
  • Input: A DNA sequence and k; optional n_bits, seed and canonical settings.
  • Output: Returns a BitFingerprintResult; the example shows dimension, values.

  • Common parameters: Default n_bits=2048, canonical=True, seed=0; representation="sparse" only saves bits with value 1.

  • Example code:

from dnakit import DNASequence
from dnakit.fingerprints import hashed_kmer_fingerprint

result = hashed_kmer_fingerprint(
    DNASequence("ACGTAC"),
    k=3,
    n_bits=16,
    seed=7,
    representation="sparse",
)

print(result.dimension)
print(dict(result.values))
  • Example output:
16
{'bit:1': 1, 'bit:10': 1}

2) Panel existence fingerprint

  • Purpose: Convert the user-defined naming pattern panel into an interpretable 0/1 bit vector. Each bit indicates whether the corresponding motif or recognition sequence exists, making it easy to compare and filter sequences according to known functional patterns.
  • Method: Scan named patterns in fixed panel order; set each bit to 1 when at least one match exists, otherwise 0.
  • API: dnakit.fingerprints.panel_fingerprint(value[required], panel[required], mode[optional], overlapping[optional], representation[optional], max_panel_size[optional], max_matches_per_pattern[optional])
  • Input: A DNA sequence and a named pattern panel with fixed order.
  • Output: Returns a BitFingerprintResult; the example shows feature_names, dense_values.

  • Mode: Default is mode="iupac", mode="exact" can also be used; by default, both positive and negative chains are scanned.

  • Example code:

from dnakit import DNASequence
from dnakit.fingerprints import panel_fingerprint

panel = {
    "start": "ATG",
    "EcoRI": "GAATTC",
    "TATA_box": "TATAWAWR",
}
result = panel_fingerprint(DNASequence("ATGGAATTC"), panel)

print(result.feature_names)
print(result.dense_values())
  • Example output:
('panel:EcoRI', 'panel:TATA_box', 'panel:start')
(1, 0, 1)
  • Explanation: The corresponding modes of EcoRI and start exist, so the corresponding bits are 1; TATA_box does not exist, so it is 0.

FAQ: details