Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Multi-modal transformer for signal classification in nanopore blockade experiments

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read By jointly processing the raw ionic-current trace, a wavelet image, and a statistical descriptor vector of the same nanopore blockade event in one transformer, this paper reports 92.6% macro-averaged accuracy across 42 peptide classes — mor

desk verdict A credible empirical win for fusing raw, wavelet, and catch22 features in nanopore classification, but the '>10pp' headline rests on baseline numbers that should have been recomputed on the same split. read the letter →

arxiv 2607.20323 v1 pith:R2DGK3JD submitted 2026-07-22 cs.LG physics.comp-phq-bio.BM

Sandro Kuppel , Julian Ho{ss}bach , Samuel Tovey , Christian Holm This is my paper · ORCID
classification cs.LGphysics.comp-phq-bio.BM
keywords nanoporesensingpeptideclassificationmulti-modaltransformercross-attentionwavelettransformcatch22featurestransferlearningsingle-moleculedetection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a single deep-learning model which processes several representations of the same nanopore blockade event at once can classify peptides far more accurately and uniformly than any method that looks at one representation alone. On a 42-peptide benchmark, the multi-modal transformer reaches 92.6% macro-averaged accuracy, more than 10 percentage points above the best single-modality baseline, and it lifts the worst-performing class from 58.7% to 77.2%. The same architecture, pre-trained on the large benchmark and fine-tuned on a smaller 20-amino-acid dataset, reaches 99.0% macro accuracy and converges faster than training from scratch. If these results hold, they would make nanopore single-molecule sensing substantially more reliable for diagnostics, where the least-well-resolved analyte limits the whole assay, and suggest that pre-trained models can reduce the cost of applying machine learning to new nanopore analytes.

What carries the argument

The key mechanism is a multi-branch transformer encoder built on the Vision Transformer. Each modality forms its own branch: the raw current trace is split into segments, the wavelet image into patches, and the catch22 descriptor vector is passed through an MLP to create a classification token. Cross-attention transfers information between branches — once from the descriptor token to the time-series and image branches before the encoder, and then alternately with standard transformer blocks inside the encoder. Masked-autoencoder pretraining, with classification tokens included and different masking ratios per branch (50% of time-series tokens, 70% of wavelet tokens), forces the branches to s

What would settle it

Recompute the ResNet18-wavelet and catch22-MLP baselines on the exact test split and preprocessing used for the multi-modal model. If the multi-modal model no longer leads by 10+ percentage points, the headline margin is not reproducible; additionally, re-labeling a sample of the test set to remove the estimated 3–5% label noise and re-evaluating would test whether the model is truly at the attainable ceiling.

Watch

Extended reading notes

Core claim

The central claim is that combining three complementary views of the same blockade current signal — the raw time series, a wavelet-transformed image, and a vector of catch22 statistical descriptors — within one multi-branch transformer yields qualitatively better peptide classification than any single view. The model achieves 92.6% macro / 92.1% micro accuracy on 42 peptide classes from a peptide-ladder experiment, versus 81.7%/81.5% for the best prior single-modality model; its worst-class accuracy jumps from 58.7% to 77.2%, and the best class reaches 100%. Attention analysis shows the representations emphasize different structures — entry/exit phases and deep short blockades for the time s

Load-bearing premise

The headline 'more than 10 percentage points' improvement rests on the assumption that the published baseline accuracies were measured on the same test split and with the same preprocessing as the multi-modal model; the paper does not state that the baselines were recomputed on identical held-out events.

Editorial extensions

If this is right

  • If the 92.6% macro accuracy is correct, the model operates near the ceiling set by the estimated 3–5% label noise, so further gains would require cleaner labels, not better architectures.
  • Worst-class accuracy of 77.2% (up from 58.7%) means a clinical assay built on this classifier would not be dragged down by a single poorly resolved peptide.
  • Transfer to the 20-amino-acid set at 99.0% macro accuracy, with faster convergence after pretraining, implies that nanopore signal features generalize across analytes and that pretraining can cut the labeled-data cost for new sensing tasks.
  • Because the architecture accepts any number and type of branches, joint analysis with additional readouts (for instance optical signals recorded alongside ionic current) can be added without redesigning the model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Both datasets come from the same aerolysin pore, so the paper demonstrates transfer across analytes, not across pore types; testing on a structurally different pore would show whether the model is a general nanopore foundation model.
  • The label-noise ceiling (3–5%) is estimated, not measured; generating a cleanly labeled subset of the peptide-ladder test set and re-evaluating would reveal whether the model's apparent ceiling is real or an artifact of noisy labels.
  • The attention patterns are observational; a targeted perturbation — e.g., zeroing the high-frequency band or the entry/exit phases during inference — would directly test whether those features are causally responsible for the wavelet and time-series branches' contributions.
  • The multi-branch design is agnostic to input type, so joint electro-optical readouts or other time-series derived representations can be plugged in; one could even imagine using the same architecture for classification of other single-molecule translocation signals.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a multi-modal transformer for classifying nanopore blockade events, combining three input representations: raw ionic-current time series, wavelet-transform images, and a small set of catch22-derived static features. The model uses branch-specific transformer encoders with cross-attention, plus a masked-autoencoder pretraining stage. On a 42-peptide ladder dataset (350,000 events), the authors report 92.6% macro / 92.1% micro accuracy, improving the best published baseline by more than 10 percentage points, and raising the worst-class accuracy from 58.7% to 77.2%. A transfer experiment to a 20-amino-acid XR7 dataset reports 99.0% macro accuracy when fine-tuning from the peptide-ladder model, with faster convergence and a narrower per-class accuracy distribution than training from scratch. Attention analysis is used to argue that time-series and wavelet modalities focus on complementary signal attributes.

Significance. If the results are reproducible, this would be a substantial advance for nanopore-based protein/peptide identification and for multi-modal learning on single-molecule signals. The architecture is generic and could be extended to additional input modalities, and the transfer result across analyte sets with the same pore type is practically important. The paper is explicit about its limitations (same pore architecture, only 42 peptide classes) and the authors provide a reasonably detailed appendix on model and training specifications. However, the central benchmark comparison currently rests on unverified baseline comparability, the headline improvement lacks uncertainty quantification, and the catch22 feature selection procedure appears to use the full dataset. The manuscript's code/data availability statement is also deferred until publication, which hampers verification.

major comments (4)
  1. [§4, Table 1] Table 1 labels the ResNet18-Wavelet and MLP-Catch22 rows as 'comparison to prior work,' but the text does not state whether these numbers were recomputed on the same 70/15/15 split and identical preprocessing pipeline used for the multi-modal model. The headline 'more than 10 percentage points' is exactly the margin between 92.6% and 81.7%; if the prior numbers used a different split or wavelet/catch22 parameters, the margin could shrink or change direction. The authors should either rerun the baselines on the exact same test events and preprocessing, or explicitly report the original settings and justify comparability.
  2. [§2, Table 1 and §4 (Model training)] No error bars, confidence intervals, or multiple seeds are reported for any accuracy number. The claim of a 10.9pp macro improvement (and the 18.5pp worst-class improvement) is the central result, yet the run-to-run variance of transformer training is often substantial. At minimum, 3–5 independent seeds should be run and mean±std reported for each model, with per-class bootstrap intervals if possible. Without this, the magnitude of the improvement is not statistically grounded.
  3. [§4, 'Experimental data and model input'] The manuscript states: 'To remove redundant data, we reduced this set to the five most relevant features identified in [17] by performing SHAP analysis on the same peptide ladder dataset used in this work.' This indicates the catch22 feature subset was selected using the full dataset, including the test partition, and before model training. This is a selection leak that can bias the reported accuracy optimistically. It is also an unfair advantage over the MLP-Catch22 baseline if that baseline uses the full feature set. The feature selection must be performed inside the training folds, or justified as using only training labels and samples.
  4. [§4, 'Experimental data and model input'; §3 Discussion] The paper states that 'Approximately 3–5% of these events are mislabeled due to limitations in the labeling process' and later uses this to argue that the multi-modal model 'approaches the ceiling attainable on this data set.' No method, reference, or measurement is given for this estimate. If the mislabel rate is different or if it is not independently estimated, the 'near ceiling' claim is unsupported. Please provide the estimation procedure or rephrase the claim as a qualitative speculation.
minor comments (5)
  1. [Abstract] 'More than 10 percentage points' is a fine summary, but the actual margins are 10.9pp macro and 10.6pp micro; given the baseline-comparability concern, the wording is somewhat optimistic. Suggest 'about 10 percentage points' or a qualified statement until baselines are recomputed.
  2. [§4, Equation (2)] The expression 'p D2/h' should read 'sqrt(D2/h)' or be typeset properly; the current inline notation is ambiguous.
  3. [§4, 'Wavelet images'] Typo: 'thehhhatwavelet' should presumably be 'the hhhat wavelet' (or 'the HHHat wavelet').
  4. [Figure 5] The 'adjusted validation loss' subtracts an 'irreducible offset' from label smoothing, but the offset value and its calculation are not specified. Please state how it was computed so the plot is reproducible.
  5. [Data and code availability] Availability 'upon publication' is standard, but for review purposes the withholding of code/data makes it impossible to verify the central benchmark. Please provide reviewer access or a detailed model/config file in the supplement.

Circularity Check

1 steps flagged · score 3.0 of 10

Catch22 feature selection on the evaluation dataset is a partial circular step; the core benchmark claim is otherwise empirical.

  1. fitted input called prediction [Section 4, 'Experimental data and model input']
    "To remove redundant data, we reduced this set to the five most relevant features identified in [17] by performing SHAP analysis on the same peptide ladder dataset used in this work. These five features form the third input to our model."

    The catch22 descriptor subset is selected by SHAP analysis on the same peptide-ladder dataset that is later split into train/validation/test and used to report the model's test accuracy; the paper does not state that the selection was restricted to the training split. The test labels have therefore already influenced which features enter the model, so the reported 92.6% macro accuracy is not a clean out-of-sample estimate for the descriptor modality. This is a selection leak: the input representation is fitted to the evaluation data and then the model is presented as predicting classes on that same dataset.

full rationale

The paper's central claim is an empirical benchmark result, not a derivation; the multi-modal model's accuracy is measured on a held-out split and is not equal to any fitted parameter by construction. The one concrete circularity-adjacent step is the catch22 feature subset: the text says the five features were chosen by SHAP analysis on the same peptide-ladder dataset used for evaluation. Unless that selection was performed on the training split alone (which is not stated), the test labels have influenced the model's input representation, making the descriptor modality partially fitted to the test set and the reported accuracy optimistic. The Table 1 baselines [30,17] are from the same research group and the paper does not demonstrate same-split recomputation, which is a load-bearing comparability/self-citation risk for the '>10 percentage points' headline, but it is a correctness concern rather than a constructional circularity. No equations reduce to their inputs and no theorem is imported by self-citation, so the overall circularity is partial rather than fundamental.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. Its free parameters are model and preprocessing choices, some of which (catch22 feature selection, label noise estimate) are tied to the same data used for evaluation. The main assumptions concern data quality, comparability of baselines, and the i.i.d. split.

free parameters (4)
  • Catch22 five-feature subset = Five features selected via SHAP on the peptide-ladder dataset (from ref [17])
    The catch22 input was reduced to five features chosen by SHAP analysis on the same peptide-ladder dataset used for evaluation, potentially leaking test information into model selection.
  • Architecture hyperparameters = D=256, hidden=1024, heads=8, blocks=12, patch=14, seq=4, MLP(128,256)
    Hand-chosen transformer hyperparameters; standard for such models but not derived from first principles.
  • Pretraining masking ratios = 50% time-series tokens, 70% wavelet tokens
    Chosen by hand; affects the quality of self-supervised pretraining and downstream accuracy.
  • Label noise estimate = 3-5%
    An informal estimate used to argue that 92.6% accuracy approaches the ceiling; no direct measurement is provided.
assumptions (4)
  • domain assumption The 3-5% label noise estimate bounds the achievable classification accuracy.
    Used in the Discussion to claim 92.6% is near the ceiling; the estimate is not measured in this paper.
  • domain assumption Baseline results from refs [17,30] are comparable to the authors' models on the same test split.
    Table 1 compares against previous methods, but the paper does not state that they were re-evaluated on the identical held-out split and preprocessing pipeline.
  • domain assumption The catch22 feature subset selected via SHAP on the same dataset does not leak test information.
    The five catch22 features were identified using the same peptide-ladder dataset, so the selection may have used test labels.
  • domain assumption The train/validation/test split is i.i.d. and events are independent samples.
    The evaluation treats each blockade event as an independent sample from a fixed distribution; no temporal or experimental batch structure is modeled.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-modal transformer for signal classification in nanopore blockade experiments." pith.science (2026). https://pith.science/paper/R2DGK3JD

@misc{pith2026260720323,
  author       = {Pith},
  title        = {Pith review of: Multi-modal transformer for signal classification in nanopore blockade experiments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R2DGK3JD}},
  note         = {Machine review of arXiv:2607.20323}
}
read the original abstract

Nanopore devices have emerged as powerful tools for single-molecule sensing, with potential for rapid, portable diagnostics. They detect changes in ionic current as analytes enter nanometer-scale pores, providing a means of identifying diverse biomarkers from their characteristic signal patterns. However, these signals are highly complex, and reliably assigning them to specific molecules remains a major challenge. Here, we address this by introducing a multi-modal deep learning architecture that jointly processes multiple signal representations, including raw time-series data, wavelet-based images, and static feature vectors. Our approach surpasses existing methods by more than 10 percentage points on a 42-peptide benchmark and transfers to a 20-amino-acid dataset with near-perfect accuracy. The model integrates complementary information from these representations, with attention analysis showing that the time-series and wavelet-image inputs emphasize different features of the same event. Together, these results demonstrate the potential of machine learning to enable robust, high-accuracy molecular identification with nanopore sensors.

Figures

Figures reproduced from arXiv: 2607.20323 by the authors.

Figure 1
Figure 1. Experimental workflow and model architecture. (A) Illustration of the experimental setup using an aerolysin nanopore embedded in a lipid bilayer over which an electric current is applied. (B) Measuring the electric current through the pore results in a continuous current readout with distinctive blockade currents, corresponding to peptides being present in the pore. These events are extracted and analyzed. (C) For c… view at source ↗
Figure 2
Figure 2. Per-class accuracy distributions on the 42-peptide ladder dataset. Dis￾tribution of per-class accuracies for all four models – ResNet18, catch22, the time-series transformer, and the multi-modal transformer – evaluated on the 42- peptide ladder dataset. The multi-modal transformer concentrates near perfect accuracy, whereas the single-modality models show broader distributions with a tail of poorly classified classe… view at source ↗
Figure 3
Figure 3. Fusion evaluation. To evaluate our model’s ability to integrate information from all three modalities, we first identify the classes where each modality per￾forms best. Specifically, we select the top 10 classes with the highest accuracies for each modality using their respective best-performing model. For the catch22 data, we use the dense neural network from [17], for the wavelet data, we show the ResNet18 from [3… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Attention analysis of a representative blockade event. Attention roll￾out [1] for the Time Series Transformer (left) and the Vision Transformer (right), shown for the same event in its raw time-series and wavelet representation. For the time series, attention weights a…
Figure 5
Figure 5. Figure 5: Transfer learning evaluation. Validation loss and accuracy during training (left) for a model trained on the XR7 dataset alone (standard) and one pre-trained on the peptide-ladder dataset and fine-tuned on XR7 (transferred). Because label smoothing prevents the cross-e…
Figure 6
Figure 6. Figure 6: Properties of the two datasets. Top row, peptide-ladder dataset; bottom row, XR7 dataset. (A) Event counts per class for the peptide ladder, highlighting the strong class imbalance. (B) Mean-current histograms for all six peptide ladders. (C) Event counts per class for…
Figure 7
Figure 7. Figure 7: Pre-training Framework. During pretraining, the model is trained to recon￾struct time series and image data from small portions of unmasked input tokens. This is achieved using an encoder-decoder architecture. The encoder processes the unmasked patches to generate a la…
Figure 8
Figure 8. Figure 8: Extended attention analysis of the vision transformer and the time series transformer. For the vision transformer, the wavelet images are displayed in grayscale, with attention weights shown in the foreground, where red represents the highest values. For the time serie…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 6 linked inside Pith

  1. [17]

    Behrends, and Christian Holm

    Julian Hoßbach, Samuel Tovey, Tobias Ensslen, Jan C. Behrends, and Christian Holm. Peptide classification from statistical analysis of nanopore sensing experiments.The Journal of Chemical Physics, 162(8):084107, February 2025

    2025

  2. [1]

    Quantifying Attention Flow in Transformers

    Samira Abnar and Willem Zuidema. Quantifying Attention Flow in Transformers. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguis- tics, pages 4190–4197, Online, July 2020. Association for Computational Linguistics

    2020

  3. [2]

    Howard, Xander F

    Javier Antonio Alfaro, Peggy Bohl¨ ander, Mingjie Dai, Mike Filius, Cecil J. Howard, Xander F. van Kooten, Shilo Ohayon, Adam Pomorski, Sonja Schmid, Aleksei Ak- simentiev, Eric V. Anslyn, Georges Bedran, Chan Cao, Mauro Chinappi, Etienne Coyaud, Cees Dekker, Gunnar Dittmar, Nicholas Drachman, Rienk Eelkema, David Goodlett, S´ ebastien Hentz, Umesh Kalath...

    2021

  4. [3]

    Kasianowicz, Manuela Pastoriza-Gallego, J´ erˆ ome Math´ e, R´ egis Daniel, Fabien Piguet, and Abdelghani Oukhaled

    Mazdak Afshar Bakshloo, John J. Kasianowicz, Manuela Pastoriza-Gallego, J´ erˆ ome Math´ e, R´ egis Daniel, Fabien Piguet, and Abdelghani Oukhaled. Nanopore-Based Protein Identification.Journal of the American Chemical Society, 144(6):2716–2725, February 2022. 18 Multi-Modal Signal Classification

    2022

  5. [4]

    Behrends and Tobias Ensslen

    Jan C. Behrends and Tobias Ensslen. Method and systems for identifying a sequence of monomer units of a biological or synthetic heteropolymer, Patent US20240077491A1, March 7, 2024

    2024

  6. [5]

    PhD thesis, ´Ecole doctorale Sciences et ing´ enierie (Cergy-Pontoise, Val d’Oise), 2018

    Mordjane Boukhet.Discrimination and Sequencing of Polymers with Biological Nanopores. PhD thesis, ´Ecole doctorale Sciences et ing´ enierie (Cergy-Pontoise, Val d’Oise), 2018. 2018CERG0984

    2018

  7. [6]

    CrossViT: Cross-attention multi-scale vision transformer for image classification

    Chun-Fu Richard Chen, Quanfu Fan, and Rameswar Panda. CrossViT: Cross-attention multi-scale vision transformer for image classification. In2021 IEEE/CVF Interna- tional Conference on Computer Vision (ICCV), pages 347–356, Montreal, QC, Canada, October 2021. IEEE

    2021

  8. [7]

    Vision Trans- formers Need Registers

    Timoth´ ee Darcet, Maxime Oquab, Julien Mairal, and Piotr Bojanowski. Vision Trans- formers Need Registers. https://arxiv.org/abs/2309.16588, October 2025

    arXiv 2025

Show all 38 references
  1. [8]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. https://arxiv.org/abs/1810.04805, May 2019

    2019 arXiv

  2. [9]

    An Image Is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An Image Is Worth 16x16 Words: Transformers for Image Recognition at...

    2010 arXiv

  3. [10]

    PhD thesis, Albert-Ludwigs-Universit¨ at Freiburg, Freiburg, 2024

    Tobias Ensslen.Mechanistic Principles of High- Resolution Discrimination of Polymers with Membrane Proteins. PhD thesis, Albert-Ludwigs-Universit¨ at Freiburg, Freiburg, 2024

    2024

  4. [11]

    Behrends

    Tobias Ensslen, Kumar Sarthak, Aleksei Aksimentiev, and Jan C. Behrends. Resolving Isomeric Posttranslational Modifications Using a Biological Nanopore as a Sensor of Molecular Shape.Journal of the American Chemical Society, 144(35):16060–16068, September 2022

    2022

  5. [12]

    Behrends

    Tobias Ensslen, Jessica Schwaiger, and Jan C. Behrends. In situ sequence recognition of peptides by enzymatic cleavage.Biophysical Journal, 123(3):146a, February 2024

    2024

  6. [13]

    Dumitru Erhan, Aaron Courville, Yoshua Bengio, and Pascal Vincent. Why does unsupervised pre-training help deep learning? InProceedings of the Thirteenth Inter- national Conference on Artificial Intelligence and Statistics, volume 9 ofProceedings of Machine Learning Research, ...

    2010

  7. [14]

    Fried, Yanfang Wu, Richard D

    Jasper P. Fried, Yanfang Wu, Richard D. Tilley, and J. Justin Gooding. Optical Nanopore Sensors for Quantitative Analysis.Nano Letters, 22(3):869–880, February 2022. 19 Kuppel et al

    2022

  8. [15]

    Burman, Javier E

    Joseph Hart, Oliver P. Burman, Javier E. L. Villa, Yilin Li, Kyan Li, Wei Xin Chua, Is- abel Garc ´ ıa, Ren Ren, Sarah A. Goodchild, Luis M. Liz-Marz´ an, Aleksandar P. Ivanov, and Joshua B. Edel. NanoBoost: a wavelet transform-enhanced machine learning al- gorithm for nanopor...

    2026

  9. [16]

    Masked Autoencoders Are Scalable Vision Learners.2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15979–15988, June 2022

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ ar, and Ross Girshick. Masked Autoencoders Are Scalable Vision Learners.2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15979–15988, June 2022

    2022

  10. [18]

    Finishing the euchromatic se- quence of the human genome.Nature, 431(7011):931–945, October 2004

    International Human Genome Sequencing Consortium. Finishing the euchromatic se- quence of the human genome.Nature, 431(7011):931–945, October 2004

    2004

  11. [19]

    SGDR: Stochastic Gradient Descent with Warm Restarts

    Ilya Loshchilov and Frank Hutter. SGDR: Stochastic Gradient Descent with Warm Restarts. https://arxiv.org/abs/1608.03983, May 2017

    2017 arXiv

  12. [20]

    Lubba, Sarab S

    Carl H. Lubba, Sarab S. Sethi, Philip Knaute, Simon R. Schultz, Ben D. Fulcher, and Nick S. Jones. Catch22: CAnonical Time-series CHaracteristics. https://arxiv.org/abs/1901.10200, January 2019

    1901 arXiv

  13. [21]

    Karolis Misiunas, Niklas Ermann, and Ulrich F. Keyser. QuipuNet: Convolutional Neural Network for Single-Molecule Nanopore Sensing.Nano Letters, 18(6):4040–4045, June 2018

    2018

  14. [22]

    Fox, Yishu Fang, Melissa Queen, Mattias Tolhurst, Nicolas Cardozo, Miten Jain, and Jeff Nivala

    Keisuke Motone, Daphne Kontogiorgos-Heintz, Jasmine Wee, Kyoko Kurihara, Sang- beom Yang, Gwendolin Roote, Oren E. Fox, Yishu Fang, Melissa Queen, Mattias Tolhurst, Nicolas Cardozo, Miten Jain, and Jeff Nivala. Multi-pass, single-molecule nanopore reading of long protein stran...

    2024

  15. [23]

    Ssqueezepy.GitHub

    John Muradeli. Ssqueezepy.GitHub. Note: https://github.com/OverLordGoldDragon/ssqueezepy/, 2020

    2020

  16. [24]

    Behrends, Aleksei Aksimentiev, and Abdelghani Oukhaled

    Hadjer Ouldali, Kumar Sarthak, Tobias Ensslen, Fabien Piguet, Philippe Manivet, Juan Pelta, Jan C. Behrends, Aleksei Aksimentiev, and Abdelghani Oukhaled. Elec- trical recognition of the twenty proteinogenic amino acids using an aerolysin nanopore. Nature Biotechnology, 38(2):...

    2020

  17. [25]

    Nanopore sensing of protein and peptide conformation for point-of-care applications

    Laura Ratinho, Nathan Meyer, Sandra Greive, Benjamin Cressiot, and Juan Pelta. Nanopore sensing of protein and peptide conformation for point-of-care applications. Nature Communications, 16(1):3211, April 2025

    2025

  18. [26]

    Single-Molecule Fingerprinting of Unlabeled Full-Length Proteins Using an Aerolysin Nanopore.Journal of the American Chemical Society, June 2026

    Verena Rukes, Evita Norkute, Georges Barnikol, Jingze Duan, Jiajie Gao, and Chan Cao. Single-Molecule Fingerprinting of Unlabeled Full-Length Proteins Using an Aerolysin Nanopore.Journal of the American Chemical Society, June 2026. 20 Multi-Modal Signal Classification

    2026

  19. [27]

    Knappe, Mark Bathe, George Barbastathis, and Patrick S

    Arjav Shah, Xin Kai Lee, Kun Li, Grant A. Knappe, Mark Bathe, George Barbastathis, and Patrick S. Doyle. Interpretable deep learning for single-molecule nanopore finger- printing using physics-guided preprocessing.ACS Sensors, 11(5):3828–3839, 2026

    2026

  20. [28]

    Jayne Wallace, Marjetka Podobnik, and Gregor Anderluh

    Marija Srnko, Gaˇ sperˇSolinc, Ana Crnkovi´ c, Franci Merzel, Michael Jordan, E. Jayne Wallace, Marjetka Podobnik, and Gregor Anderluh. High-throughput human histone detection by an engineered actinoporin nanopore.ACS Sensors, 11(3):2016–2029, 2026

    2016

  21. [29]

    Pre-training of Lightweight Vision Transformers on Small Datasets with Minimally Scaled Images

    Jen Hong Tan. Pre-training of Lightweight Vision Transformers on Small Datasets with Minimally Scaled Images. https://arxiv.org/abs/2402.03752, February 2024

    2024 arXiv

  22. [30]

    Behrends, and Christian Holm

    Samuel Tovey, Julian Hoßbach, Sandro Kuppel, Tobias Ensslen, Jan C. Behrends, and Christian Holm. Deep Learning-Driven Peptide Classification in Biological Nanopores. https://arxiv.org/abs/2509.14029, October 2025

    2025

  23. [31]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. InProceedings of the 31st International Conference on Neural Information Processing Systems, NIPS’17, pages 6000–6010, Red H...

    2017

  24. [32]

    Dynamic Features Driven by Stochastic Collisions in a Nanopore for Precise Single-Molecule Identification.Journal of the American Chemical Society, 147(2):1781–1791, January 2025

    Jia Wang, Shao-Chuang Liu, Zheng-Li Hu, Yi-Lun Ying, and Yi-Tao Long. Dynamic Features Driven by Stochastic Collisions in a Nanopore for Precise Single-Molecule Identification.Journal of the American Chemical Society, 147(2):1781–1791, January 2025

    2025

  25. [33]

    Reiner, Celeste Kennard, Mark J

    Xiaojun Wei, Tadas Penkauskas, Joseph E. Reiner, Celeste Kennard, Mark J. Uline, Qian Wang, Sheng Li, Aleksei Aksimentiev, Joseph W. F. Robertson, and Chang Liu. Engineering Biological Nanopore Approaches toward Protein Sequencing.ACS Nano, 17(17):16369–16395, September 2023

    2023

  26. [34]

    LIMU-BERT: Unleashing the Potential of Unlabeled Data for IMU Sensing Applications

    Huatao Xu, Pengfei Zhou, Rui Tan, Mo Li, and Guobin Shen. LIMU-BERT: Unleashing the Potential of Unlabeled Data for IMU Sensing Applications. InProceedings of the 19th ACM Conference on Embedded Networked Sensor Systems, SenSys ’21, pages 220–233, New York, NY, USA, November 2...

    2021

  27. [35]

    Single-Molecule Ionic and Optical Sensing with Nanoapertures

    Wayne Yang and Cees Dekker. Single-Molecule Ionic and Optical Sensing with Nanoapertures. InNanostructure Science and Technology, pages 367–387. Springer International Publishing, Cham, 2022

    2022

  28. [36]

    Foster, Aleksei Aksimentiev, Min Chen, and Meni Wanunu

    Luning Yu, Xinqi Kang, Fanjun Li, Behzad Mehrafrooz, Amr Makhamreh, Ali Fallahi, Joshua C. Foster, Aleksei Aksimentiev, Min Chen, and Meni Wanunu. Unidirectional single-file transport of full-length proteins through a nanopore.Nature Biotechnology, 41(8):1130–1139, August 2023

    2023

  29. [37]

    Yuan Yuan and Lei Lin. Self-Supervised Pretraining of Transformers for Satellite Image Time Series Classification.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14:474–487, 2021. 21 Kuppel et al

    2021

  30. [38]

    Peptide sequencing based on host–guest interaction-assisted nanopore sensing.Nature Methods, 21(1):102– 109, January 2024

    Yun Zhang, Yakun Yi, Ziyi Li, Ke Zhou, Lei Liu, and Hai-Chen Wu. Peptide sequencing based on host–guest interaction-assisted nanopore sensing.Nature Methods, 21(1):102– 109, January 2024. 22

    2024

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.