Nanopore Data Analysis: Baseline Construction and Abrupt Change-Based Multilevel Fitting

Abstract

Solid-state nanopore technology delivers single-molecule resolution information, and the quality of the deliverables hinges on the capability of the analysis platform to extract maximum possible events and fit them appropriately. In this work, we present an analysis platform with four baseline fitting methods adaptive to a wide range of nanopore traces (including those with a step or abrupt changes where pre-existing platforms fail) to maximize extractable events (2× improvement in some cases) and multilevel event fitting capability. The baseline fitting methods, in the increasing order of robustness and computational cost, include arithmetic mean, linear fit, Gaussian smoothing, and Gaussian smoothing and regressed mixing. The performance was tested with ultra-stable to vigorously fluctuating current profiles, and the event count increased with increasing fitting robustness prominently for vigorously fluctuating profiles. Turning points of events were clustered using the dbscan method, followed by segmentation into preliminary levels based on abrupt changes in the signal level, which were then iteratively refined to deduce the final levels of the event. Finally, we show the utility of clustering for multilevel DNA data analysis, followed by the assessment of protein translocation profiles.

Combining dynamic Monte Carlo with machine learning to study nanoparticle translocation

“…5D. While many software packages such as OpenNanopore, 85 EventPro, 86 and PyPore use the cumulative sum algorithm (CUSUM) to call events and subsequently fit them, we opted to use a fit to a Gaussian function because of its simplicity and the similarity in shape to our events. The event in Fig.…”

Over Two Million Protein Translocations Reveal Optimal Conditions for High Bandwidth Recordings, Studying Transport Energetics, and Unfolding

“…For the more ubiquitous LPF setting (i.e., 10 kHz), 2 T r is ~66 μs (the full-width half maximum (FWMH) approach allows to reach calculated durations as low as ~40 μs with reasonable errors). 16, 20, 24 For the 100 kHz LP filter, the theoretical 2 T r is ~6.6 μs (SI Figure S2). The translocation time distributions were then fitted with the first passage model: where D, l , and ν are the diffusion coefficient, effective sensing length, and drift velocity of the protein (see SI section 3 for more details).…”

Nanopore Current Enhancements Lack Protein Charge Dependence and Elucidate Maximum Unfolding at Protein’s Isoelectric Point

“…Data were acquired at 250 kHz and filtered at 5 kHz using the inbuilt Bessel low-pass filter before being digitized by Digidata 1550 B (Molecular Devices LLC). All resulting events were extracted using the EventPro platform …”