Nanopore Event-Transduction Signal Stabilization for Wide pH Range under Extreme Chaotrope Conditions.

Operation of an α-hemolysin nanopore transduction detector is found to be surprisingly robust over a critical range of pH (6-9), including physiological pH = 7.4 and polymerase chain reaction (PCR) pH = 8.4, and extreme chaotrope concentration, including 5 M urea. The engineered transducer molecule that is captured in the standard α-hemolysin nanopore detector, to transform it into a transduction detector, appears to play a central role in this stabilization process by stabilizing the channel against gating during its capture. This enables the nanopore transduction detector to operate as a single molecule "nanoscope" in a wide range of conditions, where tracking on molecular state is possible in a variety of different environmental conditions. In the case of streptavidin biosensing, results are shown for detector operation when in the presence of extreme (5 M) urea concentration. Complications involving degenerate states are encountered at higher chaotrope concentrations, but since the degeneracy is only of order two, this is easily absorbed into the classification task as in prior work. This allows useful detector operation over a wide range of conditions relevant to biochemistry, biomedical engineering, and biotechnology.

The NTD Nanoscope: potential applications and implementations.

BACKGROUND: Nanopore transduction detection (NTD) offers prospects for a number of highly sensitive and discriminative applications, including: (i) single nucleotide polymorphism (SNP) detection; (ii) targeted DNA re-sequencing; (iii) protein isoform assaying; and (iv) biosensing via antibody or aptamer coupled molecules. Nanopore event transduction involves single-molecule biophysics, engineered information flows, and nanopore cheminformatics. The NTD Nanoscope has seen limited use in the scientific community, however, due to lack of information about potential applications, and lack of availability for the device itself. Meta Logos Inc. is developing both pre-packaged device platforms and component-level (unassembled) kit platforms (the latter described here). In both cases a lipid bi-layer workstation is first established, then augmentations and operational protocols are provided to have a nanopore transduction detector. In this paper we provide an overview of the NTD Nanoscope applications and implementations. The NTD Nanoscope Kit, in particular, is a component-level reproduction of the standard NTD device used in previous research papers. RESULTS: The NTD Nanoscope method is shown to functionalize a single nanopore with a channel current modulator that is designed to transduce events, such as binding to a specific target. To expedite set-up in new lab settings, the calibration and troubleshooting for the NTD Nanoscope kit components and signal processing software, the NTD Nanoscope Kit, is designed to include a set of test buffers and control molecules based on experiments described in previous NTD papers (the model systems briefly described in what follows). The description of the Server-interfacing for advanced signal processing support is also briefly mentioned. CONCLUSIONS: SNP assaying, SNP discovery, DNA sequencing and RNA-seq methods are typically limited by the accuracy of the error rate of the enzymes involved, such as methods involving the polymerase chain reaction (PCR) enzyme. The NTD Nanoscope offers a means to obtain higher accuracy as it is a single-molecule method that does not inherently involve use of enzymes, using a functionalized nanopore instead.

Pattern recognition-informed feedback for nanopore detector cheminformatics.

Pattern recognition-informed (PRI) feedback using channel current cheminformatics (CCC) software is shown to be possible in "real-time" experimental efforts. The accuracy of the PRI classification is shown to inherit the high accuracy of our offline classifier: 99.9% accuracy in distinguishing between terminal base pairs of two DNA hairpins. The pattern recognition software consists of hidden Markov model (HMM) feature extraction software, and support vector machine (SVM) classification/ clustering software that is optimized for data acquired on a nanopore channel detection system. For general nanopore detection, the distributed HMM and SVM processing used here provides a processing speedup that allows pattern recognition to complete within the time frame of the signal acquisition - where the sampling is halted if the blockade signal is identified as not in the desired subset of events (or once recognized as nondiagnostic in general). We demonstrate that Nanopore Detection with PRI offers significant advantage when applied to data acquisition on antibody-antigen system, or other complex biomolecular mixtures, due to the reduction in wasted observation time on eventually rejected "junk" (nondiagnostic) signals.

Hybrid MM/SVM structural sensors for stochastic sequential data.

In this paper we present preliminary results stemming from a novel application of Markov Models and Support Vector Machines to splice site classification of Intron-Exon and Exon-Intron (5' and 3') splice sites. We present the use of Markov based statistical methods, in a log likelihood discriminator framework, to create a non-summed, fixed-length, feature vector for SVM-based classification. We also explore the use of Shannon-entropy based analysis for automated identification of minimal-size models (where smaller models have known information loss according to the specified Shannon entropy representation). We evaluate a variety of kernels and kernel parameters in the classification effort. We present results of the algorithms for splice-site datasets consisting of sequences from a variety of species for comparison.

Clustering ionic flow blockade toggles with a mixture of HMMs.

BACKGROUND: Ionic current blockade signal processing, for use in nanopore detection, offers a promising new way to analyze single molecule properties with potential implications for DNA sequencing. The alpha-Hemolysin transmembrane channel interacts with a translocating molecule in a nontrivial way, frequently evidenced by a complex ionic flow blockade pattern with readily distinguishable modes of toggling. Effective processing of such signals requires developing machine learning methods capable of learning the various blockade modes for classification and knowledge discovery purposes. Here we propose a method aimed to improve our stochastic analysis capabilities to better understand the discriminatory capabilities of the observed the nanopore channel interactions with analyte. RESULTS: We tailored our memory-sparse distributed implementation of a Mixture of Hidden Markov Models (MHMMs) to the problem of channel current blockade clustering and associated analyte classification. By using probabilistic fully connected HMM profiles as mixture components we were able to cluster the various 9 base-pair hairpin channel blockades. We obtained very high Maximum a Posteriori (MAP) classification with a mixture of 12 different channel blockade profiles, each with 4 levels, a configuration that can be computed with sufficient speed for real-time experimental feedback. MAP classification performance depends on several factors such as the number of mixture components, the number of levels in each profile, and the duration of a channel blockade event. We distribute Baum-Welch Expectation Maximization (EM) algorithms running on our model in two ways. A distributed implementation of the MHMM data processing accelerates data clustering efforts. The second, simultanteous, strategy uses an EM checkpointing algorithm to lower the memory use and efficiently distribute the bulk of EM processing in processing large data sequences (such as for the progressive sums used in the HMM parameter estimates). CONCLUSION: The proposed distributed MHMM method has many appealing properties, such as precise classification of analyte in real-time scenarios, and the ability to incorporate new domain knowledge into a flexible, easily distributable, architecture. The distributed HMM provides a feature extraction that is equivalent to that of the sequential HMM with a speedup factor approximately equal to the number of independent CPUs operating on the data. The MHMM topology learns clusters existing within data samples via distributed HMM EM learning. A Java implementation of the MHMM algorithm is available at http://logos.cs.uno.edu/~achurban.

Accumulation of GC donor splice signals in mammals.

UNLABELLED: The GT dinucleotide in the first two intron positions is the most conserved element of the U2 donor splice signals. However, in a small fraction of donor sites, GT is replaced by GC. A substantial enrichment of GC in donor sites of alternatively spliced genes has been observed previously in human, nematode and Arabidopsis, suggesting that GC signals are important for regulation of alternative splicing. We used parsimony analysis to reconstruct evolution of donor splice sites and inferred 298 GT > GC conversion events compared to 40 GC > GT conversion events in primate and rodent genomes. Thus, there was substantive accumulation of GC donor splice sites during the evolution of mammals. Accumulation of GC sites might have been driven by selection for alternative splicing. REVIEWERS: This article was reviewed by Jerzy Jurka and Anton Nekrutenko. For the full reviews, please go to the Reviewers' Reports section.

Implementing EM and Viterbi algorithms for Hidden Markov Model in linear memory.

BACKGROUND: The Baum-Welch learning procedure for Hidden Markov Models (HMMs) provides a powerful tool for tailoring HMM topologies to data for use in knowledge discovery and clustering. A linear memory procedure recently proposed by Miklós, I. and Meyer, I.M. describes a memory sparse version of the Baum-Welch algorithm with modifications to the original probabilistic table topologies to make memory use independent of sequence length (and linearly dependent on state number). The original description of the technique has some errors that we amend. We then compare the corrected implementation on a variety of data sets with conventional and checkpointing implementations. RESULTS: We provide a correct recurrence relation for the emission parameter estimate and extend it to parameter estimates of the Normal distribution. To accelerate estimation of the prior state probabilities, and decrease memory use, we reverse the originally proposed forward sweep. We describe different scaling strategies necessary in all real implementations of the algorithm to prevent underflow. In this paper we also describe our approach to a linear memory implementation of the Viterbi decoding algorithm (with linearity in the sequence length, while memory use is approximately independent of state number). We demonstrate the use of the linear memory implementation on an extended Duration Hidden Markov Model (DHMM) and on an HMM with a spike detection topology. Comparing the various implementations of the Baum-Welch procedure we find that the checkpointing algorithm produces the best overall tradeoff between memory use and speed. In cases where sequence length is very large (for Baum-Welch), or state number is very large (for Viterbi), the linear memory methods outlined may offer some utility. CONCLUSION: Our performance-optimized Java implementations of Baum-Welch algorithm are available at http://logos.cs.uno.edu/~achurban. The described method and implementations will aid sequence alignment, gene structure prediction, HMM profile training, nanopore ionic flow blockades analysis and many other domains that require efficient HMM training with EM.

Nanopore current transduction analysis of protein binding to non-terminal and terminal DNA regions: analysis of transcription factor binding, retroviral DNA terminus dynamics, and retroviral integrase-DNA binding.

BACKGROUND: Synthetic transcription factors (STFs) promise to offer a powerful new therapeutic against Cancer, AIDS, and genetic disease. Currently, 10% of drugs are of this type, including salicylate and tamoxifen. STFs that can appropriately target (and release) their transcription factor binding sites on native genomic DNA provide a means to directly influence cellular mRNA production. An effective mechanism for screening amongst transcription factor (TF) candidates would itself be highly valued, and such may be possible with nanopore cheminformatics methods. RESULTS: It is hypothesized that binding targets on channel-captured molecules, that are well away from the channel-captured region, can be monitored insofar as their binding status, or history, is concerned. The first set of experiments we perform to explore this "transduction" hypothesis involve non-terminal dsDNA binding to protein (DNA TATA box receptor binding to TBP), where we show new experimental results and application of a new cheminformatics data analysis method. In the second series of experiments to explore the transduction hypothesis we examine terminal (blunt-ended) dsDNA binding to protein. We show experimental results before and after introduction of HIV's DNA integrase to a solution of bifunctional "Y" shaped aptamers that have an HIV consensus terminus exposed for interaction. CONCLUSION: X-ray crystallographic studies have guided our understanding of DNA structure for almost a century. It is still difficult, however, to translate the sequence-directed curvature information obtained through these tools to actual systems found in solution. With a nanopore detector the sequence-dependent conformation kinetics of DNA, especially at the DNA terminus, can be studied in a new way while still in solution and on a single molecule basis.

Preliminary nanopore cheminformatics analysis of aptamer-target binding strength.

BACKGROUND: Aptamers are nucleic acids selected for their ability to bind to molecules of interest and may provide the basis for a whole new class of medicines. If the aptamer is simply a dsDNA molecule with a ssDNA overhang (a "sticky" end) then the segment of ssDNA that complements that overhang provides a known binding target with binding strength adjustable according to length of overhang. RESULTS: Two bifunctional aptamers are examined using a nanopore detector. They are chosen to provide sensitive, highly modulated, blockade signals with their captured ends, while their un-captured regions are designed to have binding moieties for complementary ssDNA targets. The bifunctional aptamers are duplex DNA on their channel-captured portion, and single-stranded DNA on their portion with binding ability. For short ssDNA, the binding is merely to the complementary strand of DNA, which is what is studied here - for 5-base and 6-base overhangs. CONCLUSION: A preliminary statistical analysis using hidden Markov models (HMMs) indicates a clear change in the blockade pattern upon binding by the single captured aptamer. This is also consistent with the hypothesis that significant conformational changes occur during the annealing binding event. In further work the objective is to simply extend this ssDNA portion to be a well-studied approximately 80 base ssDNA aptamer, joined to the same bifunctional aptamer molecular platform.

Analysis of nanopore detector measurements using Machine-Learning methods, with application to single-molecule kinetic analysis.

BACKGROUND: A nanopore detector has a nanometer-scale trans-membrane channel across which a potential difference is established, resulting in an ionic current through the channel in the pA-nA range. A distinctive channel current blockade signal is created as individually "captured" DNA molecules interact with the channel and modulate the channel's ionic current. The nanopore detector is sensitive enough that nearly identical DNA molecules can be classified with very high accuracy using machine learning techniques such as Hidden Markov Models (HMMs) and Support Vector Machines (SVMs). RESULTS: A non-standard implementation of an HMM, emission inversion, is used for improved classification. Additional features are considered for the feature vector employed by the SVM for classification as well: The addition of a single feature representing spike density is shown to notably improve classification results. Another, much larger, feature set expansion was studied (2500 additional features instead of 1), deriving from including all the HMM's transition probabilities. The expanded features can introduce redundant, noisy information (as well as diagnostic information) into the current feature set, and thus degrade classification performance. A hybrid Adaptive Boosting approach was used for feature selection to alleviate this problem. CONCLUSION: The methods shown here, for more informed feature extraction, improve both classification and provide biologists and chemists with tools for obtaining a better understanding of the kinetic properties of molecules of interest.

Duration learning for analysis of nanopore ionic current blockades.

BACKGROUND: Ionic current blockade signal processing, for use in nanopore detection, offers a promising new way to analyze single molecule properties, with potential implications for DNA sequencing. The alpha-Hemolysin transmembrane channel interacts with a translocating molecule in a nontrivial way, frequently evidenced by a complex ionic flow blockade pattern. Typically, recorded current blockade signals have several levels of blockade, with various durations, all obeying a fixed statistical profile for a given molecule. Hidden Markov Model (HMM) based duration learning experiments on artificial two-level Gaussian blockade signals helped us to identify proper modeling framework. We then apply our framework to the real multi-level DNA hairpin blockade signal. RESULTS: The identified upper level blockade state is observed with durations that are geometrically distributed (consistent with an a physical decay process for remaining in any given state). We show that mixture of convolution chains of geometrically distributed states is better for presenting multimodal long-tailed duration phenomena. Based on learned HMM profiles we are able to classify 9 base-pair DNA hairpins with accuracy up to 99.5% on signals from same-day experiments. CONCLUSION: We have demonstrated several implementations for de novo estimation of duration distribution probability density function with HMM framework and applied our model topology to the real data. The proposed design could be handy in molecular analysis based on nanopore current blockade signal.

SVM clustering.

BACKGROUND: Support Vector Machines (SVMs) provide a powerful method for classification (supervised learning). Use of SVMs for clustering (unsupervised learning) is now being considered in a number of different ways. RESULTS: An SVM-based clustering algorithm is introduced that clusters data with no a priori knowledge of input classes. The algorithm initializes by first running a binary SVM classifier against a data set with each vector in the set randomly labelled, this is repeated until an initial convergence occurs. Once this initialization step is complete, the SVM confidence parameters for classification on each of the training instances can be accessed. The lowest confidence data (e.g., the worst of the mislabelled data) then has its' labels switched to the other class label. The SVM is then re-run on the data set (with partly re-labelled data) and is guaranteed to converge in this situation since it converged previously, and now it has fewer data points to carry with mislabelling penalties. This approach appears to limit exposure to the local minima traps that can occur with other approaches. Thus, the algorithm then improves on its weakly convergent result by SVM re-training after each re-labeling on the worst of the misclassified vectors - i.e., those feature vectors with confidence factor values beyond some threshold. The repetition of the above process improves the accuracy, here a measure of separability, until there are no misclassifications. Variations on this type of clustering approach are shown. CONCLUSION: Non-parametric SVM-based clustering methods may allow for much improved performance over parametric approaches, particularly if they can be designed to inherit the strengths of their supervised SVM counterparts.

A novel, fast, HMM-with-Duration implementation - for application with a new, pattern recognition informed, nanopore detector.

BACKGROUND: Hidden Markov Models (HMMs) provide an excellent means for structure identification and feature extraction on stochastic sequential data. An HMM-with-Duration (HMMwD) is an HMM that can also exactly model the hidden-label length (recurrence) distributions - while the regular HMM will impose a best-fit geometric distribution in its modeling/representation. RESULTS: A Novel, Fast, HMM-with-Duration (HMMwD) Implementation is presented, and experimental results are shown that demonstrate its performance on two-state synthetic data designed to model Nanopore Detector Data. The HMMwD experimental results are compared to (i) the ideal model and to (ii) the conventional HMM. Its accuracy is clearly an improvement over the standard HMM, and matches that of the ideal solution in many cases where the standard HMM does not. Computationally, the new HMMwD has all the speed advantages of the conventional (simpler) HMM implementation. In preliminary work shown here, HMM feature extraction is then used to establish the first pattern recognition-informed (PRI) sampling control of a Nanopore Detector Device (on a "live" data-stream). CONCLUSION: The improved accuracy of the new HMMwD implementation, at the same order of computational cost as the standard HMM, is an important augmentation for applications in gene structure identification and channel current analysis, especially PRI sampling control, for example, where speed is essential. The PRI experiment was designed to inherit the high accuracy of the well characterized and distinctive blockades of the DNA hairpin molecules used as controls (or blockade "test-probes"). For this test set, the accuracy inherited is 99.9%.

Nanopore-based kinetics analysis of individual antibody-channel and antibody-antigen interactions.

BACKGROUND: The UNO/RIC Nanopore Detector provides a new way to study the binding and conformational changes of individual antibodies. Many critical questions regarding antibody function are still unresolved, questions that can be approached in a new way with the nanopore detector. RESULTS: We present evidence that different forms of channel blockade can be associated with the same antibody, we associate these different blockades with different orientations of "capture" of an antibody in the detector's nanometer-scale channel. We directly detect the presence of antibodies via reductions in channel current. Changes to blockade patterns upon addition of antigen suggest indirect detection of antibody/antigen binding. Similarly, DNA-hairpin anchored antibodies have been studied, where the DNA linkage is to the carboxy-terminus at the base of the antibody's Fc region, with significantly fewer types of (lengthy) capture blockades than was observed for free (un-bound) IgG antibody. The introduction of chaotropic agents and its effects on protein-protein interactions have also been observed. CONCLUSION: Nanopore-based approaches may eventually provide a direct analysis of the complex conformational "negotiations" that occur upon binding between proteins.

The alpha-hemolysin nanopore transduction detector - single-molecule binding studies and immunological screening of antibodies and aptamers.

BACKGROUND: Nanopore detection is based on observations of the ionic current threading a single, highly stable, nanometer-scale channel. The dimensions are such that small biomolecules and biopolymers (like DNA and peptides) can translocate or be captured in the channel. The identities of translocating or captured molecules can often be discerned, one from another, based on their channel blockade "signatures". There is a self-limiting aspect to a translocation-based detection mechanism: as the channel fits tighter around the translocating molecule the dynamic range of the ionic current signal is reduced. In this study, a lengthy, highly structure, high dynamic-range, molecular capture is sought as a key component of a transduction-based nanopore detection platform. RESULTS: A specialized role, or device augmentation, involving bifunctional molecules has been explored. The bifunctional molecule has one function to enter and blockade the channel in an information-rich self-modulating manner, while the other function is for binding (usually), located on a non-channel-captured portion of the molecule. Part of the bifunctional molecule is, thus, external to the channel and is free to bind or rigidly link to a larger molecule of interest. What results is an event transduction detector: molecular events are directly transduced into discernible changes in the stationary statistics of the bifunctional molecule's channel blockade. Several results are presented of nanopore-based event-transduction detection. CONCLUSION: It may be possible to directly track the bound versus unbound state of a huge variety of molecules using nanopore transduction detection.

Hidden Markov model variants and their application.

Markov statistical methods may make it possible to develop an unsupervised learning process that can automatically identify genomic structure in prokaryotes in a comprehensive way. This approach is based on mutual information, probabilistic measures, hidden Markov models, and other purely statistical inputs. This approach also provides a uniquely common ground for comparative prokaryotic genomics. The approach is an on-going effort by its nature, as a multi-pass learning process, where each round is more informed than the last, and thereby allows a shift to the more powerful methods available for supervised learning at each iteration. It is envisaged that this "bootstrap" learning process will also be useful as a knowledge discovery tool. For such an ab initio prokaryotic gene-finder to work, however, it needs a mechanism to identify critical motif structure, such as those around the start of coding or start of transcription (and then, hopefully more).For eukaryotes, even with better start-of-coding identification, parsing of eukaryotic coding regions by the HMM is still limited by the HMM's single gene assumption, as evidenced by the poor performance in alternatively spliced regions. To address these complications an approach is described to expand the states in a eukaryotic gene-predictor HMM, to operate with two layers of DNA parsing. This extension from the single layer gene prediction parse is indicated after preliminary analysis of the C. elegans alt-splice statistics. State profiles have made use of a novel hash-interpolating MM (hIMM) method. A new implementation for an HMM-with-Duration is also described, with far-reaching application to gene-structure identification and analysis of channel current blockade data.