Figure 1.
Convergent Vander- Lugt optical correlator.
O is a monochromatic source light, L1 and L2 are lenses, the scene and the filter planes hold both images to correlate and a CCD camera registers the correlation plane. The Fourier transform of the input image s(x,y) is obtained at the Filter plane where the optical correlation S(u,v)H*(u,v) takes place. The correlation result c(x,y) is captured by the CCD camera.
Figure 2.
Creation of mutated sequences.
Example of the 10% of mutation (6 bases) of a 60 bases sequence, creating a new sequence (bottom) from the original sequence (upper). The loci and the replacing bases are randomly chosen.
Figure 3.
Algorithm of the digital correlator.
The different steps of the digital correlation process are represented in the algorithm. This algorithm is executed for each pair object-scene. The output results are stored and the whole process is finally evaluated.
Figure 4.
Each nucleic acid base is represented by a specific number which represents the gray level of a pixel.
Figure 5.
The sequences are represented by successions of pixels that compose a n x m gray image. The queries or objects (left) have variable lengths while data bases or scenes (right) are represented with a fixed image dimension (100 x 100 in this example).
Figure 6.
The scene and object images are Fourier transformed. The object is filtered. Both Fourier Transforms are correlated. The peak indicates matching.
Table 1.
Correlation statistical indexes (%).
Figure 7.
The graphics presents the correlation peaks of the four objects and the four scenes presented in Figure 5. When the object sequence is correlated with the scene from where it was extracted, an outstanding peak can be observed indicating alignment (diagonal). In the other correlations, peaks are less outstanding.
Table 2.
Correlation peak relative position e.
Figure 8.
Peak and sequence length relationship
. The amplitude of the peak in the correlation plane increases when the sequence length increases; however their relationship is not lineal.
Table 3.
Positive peak mean values for several noise levels (%).
Figure 9.
Correlation with great data bases.
The figure presents the correlation result of the sequence so33 (400 bases) correlated with a scene that contains the complete database (1000×1000 bases). This shows that the size of the scenes may also vary.
Table 4.
Time (seconds) used by the main functions involved in the correlation process.
Table 5.
Sensitivity (%) of the Correlator and BLAST %.
Figure 10.
Sensitivity and mutation level.
The sensitivity of BLAST decays faster than the sensitivity of the correlation, when the percentage of mutation increases. The latter recognizes similarity with more than 40% of mutated pairs.
Table 6.
Time (seconds) taken to process 303 objects in 100 scenes.