Fast and automatic processing of multi-level events in nanopore translocation experiments Camille Raillon, Pierre Granjon, Michael Graf, Lorentz Steinbock, Aleksandra Radenovic To cite this version: Camille Raillon, Pierre Granjon, Michael Graf, Lorentz Steinbock, Aleksandra Radenovic. Fast and automatic processing of multi-level events in nanopore translocation experiments. Nanoscale, Royal Society of Chemistry, 2012, 14 (16), pp.4916-4924. 10.1039/C2NR30951C. hal-00717284 HAL Id: hal-00717284 https://hal.archives-ouvertes.fr/hal-00717284 Submitted on 22 Aug 2013 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. View Online / Journal Homepage Nanoscale Dynamic Article LinksC< Cite this: DOI: 10.1039/c2nr30951c www.rsc.org/nanoscale FEATURE ARTICLE Fast and automatic processing of multi-level events in nanopore translocation experiments† C. Raillon,a P. Granjon,b M. Graf,a L. J. Steinbocka and A. Radenovic*a Received 20th April 2012, Accepted 13th June 2012 DOI: 10.1039/c2nr30951c We have developed a method to analyze in detail, translocation events providing a novel and flexible tool for data analysis of nanopore experiments. Our program, called OpenNanopore, is based on the cumulative sums algorithm (CUSUM algorithm). This algorithm is an abrupt change detection algorithm that provides fitting of current blockages, allowing the user to easily identify the different levels in each event. Our method detects events using adaptive thresholds that adapt to low-frequency variations in the baseline. After event identification, our method uses the CUSUM algorithm to fit the levels inside every event and automatically extracts their time and amplitude information. This facilitates the statistical analysis of an event population with a given number of levels. The obtained information improves the interpretation of interactions between the molecule and nanopore. Since our program does not require any prior information about the analyzed molecules, novel molecule– nanopore interactions can be characterized. In addition our program is very fast and stable. With the progress in fabrication and control of the translocation speed, in the near future, our program could be useful in identification of the different bases of DNA. 1. Introduction Nanopores are nanometric holes in thin insulating membranes aLaboratory of Nanoscale Biology, Institute of Bioengineering, School of Engineering, EPFL, 1015 Lausanne, Switzerland. E-mail: aleksandra. existing in two modalities, protein/biological pores and solid- radenovic@epfl.ch state pores. Protein pores are made using a pore-forming protein b Published on 19 June 2012 http://pubs.rsc.org | doi:10.1039/C2NR30951C Grenoble Image Speech Signal Automatics Laboratory, Grenoble Institute 1 Downloaded by ECOLE POLYTECHNIC FED DE LAUSANNE on 11 July 2012 such as a-hemolysin that is inserted in a lipid bilayer whereas of Technology, Grenoble, France solid-state pores are fabricated in an insulating membrane using † Electronic supplementary information (ESI) available. See DOI: 2 3 10.1039/c2nr30951c highly focused ions or electrons. Camille Raillon received her Pierre Granjon received his M.Sc. in Electrical Engineering Ph.D. degree from the Grenoble with Honors in 2007 from Gre- Institute of Technology noble Institute of Technology (INPG), France in 2000. He (INPG). During her master’s, joined the Laboratoire des she participated in a student Images et des Signaux (LIS) in exchange program at the 2002 and the Gipsa-lab at INPG University of Sherbrooke, Can- in 2007, where he currently holds ada where she did her master’s a position as associate pro- thesis: fast prototyping of fessor.His current research area microfluidic channels in SU-8 mainly concerns signal process- photoresist for BioMEMS. ing for condition monitoring and After graduating from INPG, power system monitoring C: Raillon Camille worked as a character- P: Granjon (power systems, power ization engineer for a year in a networks, batteries.). He is MEMS-based pressure sensors particularly interested in linear company in Silicon Valley, California. Camille is now a Ph.D. and non-linear optimal filtering, multi-component signal analysis student at EPFL. Her Ph.D. project focuses on the single-molecule and sequential change detection algorithms. study of transcription using nanopore sensing. This journal is ª The Royal Society of Chemistry 2012 Nanoscale View Online Nanopores are used as biosensors for single-molecule detec- molecules using scatter plots, dwell time and current blockade tion; they can detect unlabeled biopolymers such as DNA and point histograms. Other studies16,17 successfully identified pop- RNA,4–8 single proteins,9,10 ligand or protein–DNA ulations amongst DNA translocations through small nanopores complexes11–13 and also RNA–antibiotic complexes.14The detec- (2–5 nm in diameter) and linked those populations with mole- tion method is simple: when a molecule passes through the cule–nanopore interactions. Dwell time histograms have also nanopore the ionic current is significantly reduced because the been used to perform molecule sorting,6 where l DNA and regular flow of ions passing through the pore is blocked. While fragments of l DNA digested by the restriction enzyme HindIII nanopore detection of those molecules has been extensively can be differentiated. This work relies on the fact that those studied and optimized, data analysis is still not standardized and DNA fragments have different lengths, hence shorter fragments can be very challenging. As a preferred analyzing tool, most of translocate faster than longer fragments. Since the speed at which the groups use time distribution of ionic current to classify the a DNA molecule translocates the pore varies significantly over hundreds of events that are collected in a single experiment. Such an experiment,18 finer analysis is required and it is typically a histogram (called a point histogram here in the text) can be performed via examination of current blockages that a trans- used to identify peaks in the current signal.15 Once those peaks locating molecule produces. Here we list examples where the have been identified, event extraction can be done with a existing method has been successfully applied: Skinner et al.19 for computer-based program using a threshold. Finally, the mean example were able to distinguish between single and double blockage and the event duration (or dwell time) can be calcu- stranded nucleic acids using point histograms of current block- lated. This method performs well and is commonly used but lacks ades in solid-state nanopores. Other groups were able to identify information on the different levels inside each event. nucleotides by immobilizing homopolymers or more complex Using this method on data displaying low noise, Meller et al.7 oligonucleotides in a-hemolysin.20,21 were able to discriminate between single polynucleotide As pointed out above, the point histogram technique performs well for current signals with low IRMS but analysis of noisier data still lacks a fast and robust data processing technique. Some commercial software solutions exist, such as pCLAMP from Michael Graf (born 1989) Molecular Devices, Inc. Although pCLAMP is intended for obtained his B.Sc. in Life acquisition and analysis of electrophysiology data it can be also Sciences and Technology from used in nanopore data analysis. On the other hand, free software EPFL in 2011. He is currently packages such as QuB exist. QuB is based on Hidden Markov finishing his M.Sc. in bioengi- Models22,23 and is intended for the analysis of generalized single- neering with specialization in molecule kinetics. Prior knowledge of the signal is required to biomedical technologies. His estimate the statistical model parameters. The group of research interests extend over S. Winters-Hilt has also reported methods to classify and cluster the areas of biophysics, infor- events using hidden Markov models (HMM).24 Those methods matics, molecular biology and give statistical models of resistivity and dwell times with rate genomics. constants and transitions between states. There are other statis- Published on 19 June 2012 http://pubs.rsc.org | doi:10.1039/C2NR30951C Downloaded by ECOLE POLYTECHNIC FED DE LAUSANNE on 11 July 2012 tical models that have been developed and applied to nanopore data, for example, classification of events using support vector M: Graf Lorenz Steinbock studied Aleksandra Radenovic received Molecular Biotechnology her master’s degree in physics (B.Sc. & M.Sc.) at the univer- from the University of Zagreb in sities in Heidelberg, Germany, 1999 before joining Professor Waterloo, Canada and Cam- Giovanni Dietler’s Laboratory bridge, UK. He graduated in of Physics of Living Matter in 2006 followed by a Ph.D. in 2000 at University of Lausanne. physics at the University of There she earned her Doctor of Cambridge, UK in Ulrich Keys- Sciences degree in 2003. In 2003 er’s group. He worked on she was also awarded a research rupture force experiments with scholarship for young DNA aptamers, pioneered the researchers