Fig 1.
Example of a locality sensitive hash (LSH), derived from a DNA sequence.
In SIG-DB, the k-mers are created using a sliding window of 1 character, as illustrated, with a sequence of length n resulting in ân-kâ k-mers.
Fig 2.
SIG-DB protocol for (1) hashing sequence into locality sensitive hash (LSH), (2) encrypting and passing LSH, (3) hashing database elements into LSHs, (4) comparing encrypted query LSH to unencrypted database LSHs, (5) passing scores and decrypting, and (6) calculating IoU, IoD, and IoQ scores.
Unencrypted space = blue boxes and encrypted space = green boxes.
Fig 3.
Intersection over union (IoU) scores for k-mer = {8, 16, 32} with random query sequence mutation rates from 0â100%.
K-mer = 8 showed best performance, with correct identification of sequence of interest up to mutation rates of 45%. The red circles indicate the largest mutation rate for each k-mer size that returned the correct result as the highest IoU value. The horizontal red line represents IoU = 0.1.
Table 1.
SIG-DB algorithm performance based on proportion of sequence randomly mutated in E. coli and S. aureus (DB = 50 seqs, LSH = 100k bases, Q = 20K bases, E = 20K bases).
Table 2.
SIG-DB algorithm performance based on mutations localized to one half of total sequence (DB = 50 seqs, LSH = 100k bases, Q = 20K bases, E = 20K bases).
Table 3.
SIG-DB runtime for increasing database sizes (Mut = 0, IoU = 1.0, K = 8).
Table 4.
SIG-DB algorithm performance based on relative sizes of query and database elements (K = 8, DB = 50 seqs, LSH = 100k bases, Mut = 0).