Skip to main content


Figure 2 | BMC Bioinformatics

Figure 2

From: The α-Hemolysin nanopore transduction detector – single-molecule binding studies and immunological screening of antibodies and aptamers

Figure 2

(a) Nanopore cheminformatics & control architecture. The figure shows the signal processing architecture that was used to classify DNA hairpins with this approach: Signal acquisition was performed using a time-domain, thresholding, Finite State Automaton. Hidden Markov Model processing with Expectation-Maximization was used for feature extraction on acquired channel blockades. Classification was then done by Support Vector Machine – the architecture for resolving the five DNA hairpin controls is shown. Four DNA hairpin control molecules have nine base-pair stem lengths that only differed in their blunt-ended DNA termini, the fifth control was an eight base-pair DNA hairpin. The accuracy shown is obtained upon completing the 15th single molecule sampling/classification (in approx. 6 seconds), where SVM-based rejection on noisy signals was employed. In recent augmentations to this architecture, a LabWindows Server is now used. Data is then sent to cluster of Linux Clients via TCP/IP channel. Linux clients run expensive HMM analysis as distributed processes (similarly for off-line SVM training). The sample classification is used by the Server to provide feedback to the nanopore apparatus to increase the effective sampling time on the molecules of interest (this can boost nanopore detector productivity by magnitudes). A test case of such sampling-control feedback is shown in [1]. (b) DNA hairpin controls and their diagnostic signals. The secondary structure of the DNA hairpins studied is shown on the right, with their highest scoring diagnostic signals shown on the left. Each signal trace start at approximately 120 pA open channel current and all blockade in a range 40–60 pA upon "capture" of the associated DNA hairpin. Even so, the signal traces have discernibly different blockade structure, which can be extracted using a Hidden Markov Model (see [8] for further details).

Back to article page