Fig 1.
Image processing removes artifacts and improves visualization.
A series of image processing algorithms is applied to the original region of interest (left) selected by the user. imfill was applied to remove any hole artifacts, followed by medfilt2 and weiner2 to remove Gaussian noise adaptively. The result is the denoised image (middle). To improve visualization for the user, adapthisteq is applied to adaptively improve the contrast of the ladder markers and sample bands from the background, generating a contrast-enhanced image (right).
Fig 2.
The lane boundaries and the locations of the Thermo Scientific GeneRuler 1 kb Plus DNA ladder markers are automatically detected by the software.
(A) All edges in the images are identified using a non-selective Canny edge detector, and the Sobel filter selectively isolates the vertical edges, identifying the lane boundaries for the ladder (red) and the sample lanes (blue). (B) Given the user-confirmed location of the ladder region m ´ n, the average intensity of the ladder across the pixel rows m is averaged over n. Peak detection finds the pixel location of each marker in m. The black triangles with corresponding base-pair values shows the peak detections for each ladder marker.
Fig 3.
Summary of the GelInsight graphical user interface.
The Image Processing tab (left) displays the input image, and the region of interest annotated with the lane boundary and ladder marker detections. The user input panel is outlined with a black dashed line. Following analysis, the software switches to the Results tab (right) where a table and several plots display the DNA fragmentation pattern analysis result for each sample in the region of interest.
Fig 4.
Peak base pair measurement accuracy using GelInsight as a function of agarose percentage.
(A) Representative ladders used in the analysis: Invitrogen 1 kb plus, Invitrogen 100 bp, New England Biolabs (NEB) 1 kb. (B) TapeStation results and peak detection errors for each band in the DNA ladders, measured from the Genomic DNA (GDNA) and the D5000 ScreenTape Assays. (C) Average GelInsight peak detection errors for each band in the sample DNA ladders across N = 3 replicates for a range of agarose percentages, from 0.50% to 2.00%. Invitrogen 1 kb plus was used as the reference ladder, while Invitrogen 100 bp and NEB 1 kb were used as the sample ladders. The asterisk (*) denotes missed peaks detections.
Fig 5.
Comparison of peak base pair detection using GelInsight with standard methods.
(A) Representative gel electrophoresis image of sonicated DNA samples, and (B) corresponding waterfall plots displaying the base-pair intensity curves for each DNA sample. (C) The average peak base pair across the DNA samples (n = 8) using GelInsight, ImageJ analysis, and TapeStation ScreenTape assays (D1000 and D5000) for each replicate (N = 3). (D) The relative peak area measurements across the DNA samples (n = 8) using GelInsight, ImageJ analysis, and TapeStation ScreenTape assays (D1000 and D5000) for each replicate (N = 3).
Fig 6.
Noisy peaks are detected by GelInsight.
Example gel electrophoresis image of a sonicated DNA sample with the reference ladder (left) and corresponding peak detections by GelInsight (right).
Fig 7.
Processing time increases with the number of samples.
Processing time as a function of the number of samples, extrapolated up to 32 samples.