pith. sign in

arxiv: 2604.06743 · v1 · submitted 2026-04-08 · ❄️ cond-mat.mes-hall · physics.data-an

Resolving Single-Peptide Phosphorylation Dynamics in Plasmonic Nanopores using Physics-Informed Bi-Path Model

Mulusew W. Yaltaye , Yingqi Zhao , Kuo Zhan , Vahid Farrahi , Jian-An Huang This is my paper

Pith reviewed 2026-05-10 17:57 UTC · model grok-4.3

classification ❄️ cond-mat.mes-hall physics.data-an
keywords single-molecule SERSplasmonic nanoporespeptide phosphorylationphysics-informed deep learningmultiple-instance learningtemporal encoderpost-translational modificationsnanopore detection
0
0 comments X

The pith

A physics-informed bi-path deep learning model identifies single-peptide phosphorylation from noisy plasmonic nanopore SERS signals.

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

The paper introduces a framework that pairs particle-in-pore plasmonic nanopores with a deep learning architecture to detect phosphorylation on individual peptides from their surface-enhanced Raman signals. Long trajectories are segmented by Pearson correlation, then fed into a temporal encoder of convolutional networks and bidirectional recurrent units inside a multiple-instance learning setup that tolerates label uncertainty and diffusion effects. A reader would care because phosphorylation reports on cellular signaling yet is difficult to observe reliably at single-molecule resolution amid background and blinking noise, so a working method would expand label-free phosphoproteomics. The approach claims to maintain reliable calls where raw spectral inspection fails.

Core claim

The central claim is that coupling nanoplasmonic confinement with a physics-informed temporal encoder that combines temporal convolutional networks and bidirectional gated recurrent units, trained under weakly supervised multiple-instance learning after Pearson-correlation segmentation of trajectories, enables robust distinction of single-peptide phosphorylation despite strong background interference and stochastic signal fluctuations.

What carries the argument

The bi-path temporal encoder, a network that merges temporal convolutional layers with bidirectional gated recurrent units inside a multiple-instance learning wrapper, which extracts both short-scale spectral variability and longer blinking dynamics while handling diffusion-driven heterogeneity through correlation-based segmentation.

If this is right

  • Single-peptide phosphorylation events can be distinguished reliably from background and fluctuations.
  • High-fidelity detection of single-molecule post-translational modifications becomes feasible without labels.
  • Ultrasensitive phosphoproteomic analysis is advanced through the combination of nanopore confinement and spatiotemporal learning.
  • The framework extends the usable range of SM-SERS to subtle chemical modifications that were previously obscured.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same segmentation-plus-weak-supervision strategy could be tested on other single-molecule spectroscopic modalities that face similar trajectory heterogeneity.
  • If the encoder generalizes across peptide sequences, it might support multiplexed detection of multiple PTM types in one experiment.
  • Hardware integration with microfluidic delivery could allow the method to process many peptides in parallel for higher throughput.

Load-bearing premise

That segmenting long trajectories by Pearson correlation and applying weakly supervised multiple-instance learning on the bi-path temporal encoder sufficiently resolves label ambiguity and diffusion-driven heterogeneity to produce reliable PTM calls.

What would settle it

A controlled test on synthetic or calibrated SM-SERS datasets with known phosphorylation states showing whether classification accuracy falls below 80 percent when the Pearson-correlation segmentation is replaced by fixed-length windows.

Figures

Figures reproduced from arXiv: 2604.06743 by Jian-An Huang, Kuo Zhan, Mulusew W. Yaltaye, Vahid Farrahi, Yingqi Zhao.

Figure 1
Figure 1. Figure 1: (A) Schematic of the plasmonic particle-in-pore sensor with a hot spot that excites part of single peptide. (B) the diffusion dynamics of molecules on the surface of gold-nanoparticles and the occupancy of the molecules the plasmonic hot spot. (C) the spectral map of single-peptide PTM to demonstrate variation in time, the color bar indicating the normalized intensity. (D) The overview of physics-informed … view at source ↗
read the original abstract

Protein phosphorylation provides a dynamic readout of cellular signaling yet remains difficult to detect at low abundance and stoichiometry. Single-molecule surface-enhanced Raman spectroscopy (SM-SERS) using particle-in-pore plasmonic nanopores offers label-free molecular detection with submolecular sensitivity. However, reliable identification of subtle post-translational modifications (PTMs) is hindered by the stochastic nature of SM-SERS signals, partial excitation of peptide residues within the plasmonic hotspot, and background interference. Here, we introduce a physics-informed deep learning framework to decode complex SM-SERS dynamics and identify single-peptide PTMs. The model integrates multiple-instance learning with a temporal encoder combining temporal convolutional networks and bidirectional gated recurrent units to capture both local spectral variability and long-range blinking dynamics. To address diffusion-driven spectral heterogeneity, long spectral trajectories are segmented using Pearson-correlation, enabling weakly supervised training under label ambiguity. This framework robustly distinguishes single peptide phosphorylation despite strong background interference and stochastic signal fluctuations. By coupling nanoplasmonic confinement with spatiotemporal deep learning, our approach enables high-fidelity detection of single-molecule phosphorylation events and advances ultrasensitive phosphoproteomic analysis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces a physics-informed deep learning framework to decode single-peptide phosphorylation dynamics from stochastic SM-SERS signals in particle-in-pore plasmonic nanopores. It integrates multiple-instance learning (MIL) with a bi-path temporal encoder (temporal convolutional networks combined with bidirectional gated recurrent units) to capture local spectral variability and long-range blinking. Long trajectories are segmented via Pearson correlation to mitigate diffusion-driven heterogeneity, enabling weakly supervised training under label ambiguity. The central claim is that this approach robustly distinguishes phosphorylated peptides despite strong background interference and signal fluctuations.

Significance. If the central claims are substantiated with quantitative validation, the work could meaningfully advance label-free single-molecule phosphoproteomics by coupling nanoplasmonic confinement with spatiotemporal deep learning. This would address key barriers in detecting low-abundance PTMs and provide a new tool for studying cellular signaling at submolecular resolution.

major comments (2)
  1. [Abstract] Abstract: The Pearson-correlation segmentation of long spectral trajectories is presented as addressing diffusion-driven heterogeneity without any validation against known diffusion models, synthetic trajectories, or physical checks (e.g., partial hotspot excitation). This step is load-bearing for the central claim because the abstract states it directly enables reliable PTM calls by resolving label ambiguity; if the metric does not isolate physically consistent segments, errors propagate into the MIL bags and undermine the robustness result.
  2. [Abstract] Abstract: No performance metrics, ablation studies, training details, or comparisons against ground-truth phosphorylation states are reported to support the claim that the framework 'robustly distinguishes' PTMs under background interference. This is load-bearing because the abstract positions the bi-path MIL model as resolving stochastic fluctuations and label ambiguity, yet without quantitative evidence the soundness of the distinction cannot be assessed.
minor comments (2)
  1. [Abstract] The abstract would benefit from a clearer description of how the TCN and biGRU paths are combined in the bi-path encoder and how the physics-informed aspect extends beyond segmentation.
  2. Consider adding citations to prior SM-SERS and nanopore PTM detection literature to better situate the novelty of the weakly supervised approach.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback on our manuscript. We address each major comment point by point below and have revised the abstract to incorporate additional context and quantitative highlights where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The Pearson-correlation segmentation of long spectral trajectories is presented as addressing diffusion-driven heterogeneity without any validation against known diffusion models, synthetic trajectories, or physical checks (e.g., partial hotspot excitation). This step is load-bearing for the central claim because the abstract states it directly enables reliable PTM calls by resolving label ambiguity; if the metric does not isolate physically consistent segments, errors propagate into the MIL bags and undermine the robustness result.

    Authors: We thank the referee for this observation. The Pearson correlation metric was selected because it quantifies linear spectral similarity over time, which is physically motivated by the expected behavior of diffusion-driven trajectories within the plasmonic hotspot. The full manuscript validates this segmentation using synthetic SM-SERS trajectories generated from established diffusion models (detailed in Methods and Supplementary Note 3) and includes physical consistency checks for partial hotspot excitation. To address the concern directly in the abstract, we have revised it to briefly reference these validation steps. revision: yes

  2. Referee: [Abstract] Abstract: No performance metrics, ablation studies, training details, or comparisons against ground-truth phosphorylation states are reported to support the claim that the framework 'robustly distinguishes' PTMs under background interference. This is load-bearing because the abstract positions the bi-path MIL model as resolving stochastic fluctuations and label ambiguity, yet without quantitative evidence the soundness of the distinction cannot be assessed.

    Authors: We agree that the abstract does not contain specific numerical results due to space limitations. The main text and supplementary materials provide quantitative performance metrics (e.g., accuracy and F1 scores under controlled interference), ablation studies on the temporal encoder components, training details, and comparisons to ground-truth phosphorylation states obtained from control experiments with known peptide modifications. We have revised the abstract to include key performance highlights supporting the robustness claim. revision: yes

Circularity Check

0 steps flagged

No circularity: model presented as independent framework without self-referential reductions

full rationale

The paper describes a new physics-informed deep learning framework that integrates multiple-instance learning with a bi-path temporal encoder (TCN + biGRU) and Pearson-correlation-based trajectory segmentation to handle SM-SERS signal heterogeneity. No equations, derivations, or self-citations are shown that reduce the PTM identification output to fitted inputs by construction, nor is any uniqueness theorem or ansatz imported from prior author work in a load-bearing way. The central claim of robust distinction under interference rests on the model's training and architecture rather than tautological redefinition of its own parameters or data splits. This is the common case of a self-contained modeling paper with no exhibited circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient technical detail to enumerate specific free parameters, axioms, or invented entities; the model is described at the level of architecture names only.

pith-pipeline@v0.9.0 · 5520 in / 1134 out tokens · 19464 ms · 2026-05-10T17:57:27.732138+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    Finally, a Pearson correlation algorithm was used to isolate particle-trapping events from empty nanopore noise, and the data were segmented for bi-path hierarchical modeling

    and standardized via min–max normalization. Finally, a Pearson correlation algorithm was used to isolate particle-trapping events from empty nanopore noise, and the data were segmented for bi-path hierarchical modeling. SERS Spectra Segmentation To resolve stochastic spectral fluctuations from peptide diffusion and blinking, we employed Pearson correlatio...

    work page 2025

  2. [2]

    Loss of con- sciousness reduces the stability of brain hubs and the heterogeneity of brain dy- namics

    https://doi.org/10.1038/s42003- 025-09131-3. (2) Yu, F.; Teo, G. C.; Kong, A. T.; Haynes, S. E.; Avtonomov, D. M.; Geiszler, D. J.; Nesvizhskii, A. I. Identification of Modified Peptides Using Localization-Aware Open Search. Nat. Commun. 2020, 11 (1),

    work page doi:10.1038/s42003- 2020

  3. [3]

    (3) Lancaster, N

    https://doi.org/10.1038/s41467-020-17921-y. (3) Lancaster, N. M.; Sinitcyn, P.; Forny, P.; Peters-Clarke, T. M.; Fecher, C.; Smith, A. J.; Shishkova, E.; Arrey, T. N.; Pashkova, A.; Robinson, M. L.; Arp, N.; Fan, J.; Hansen, J.; Galmozzi, A.; Serrano, L. R.; Rojas, J.; Gasch, A. P.; Westphall, M. S.; Stewart, H.; Hock, C.; Damoc, E.; Pagliarini, D. J.; Za...

    work page doi:10.1038/s41467-020-17921-y 2024

  4. [4]

    (4) Huang, J.; Mousavi, M

    https://doi.org/10.1038/s41467-024-51274-0. (4) Huang, J.; Mousavi, M. Z.; Giovannini, G.; Zhao, Y.; Hubarevich, A.; Soler, M. A.; Rocchia, W.; Garoli, D.; De Angelis, F. Multiplexed Discrimination of Single Amino Acid Residues in Polypeptides in a Single SERS Hot Spot. Angew. Chem. Int. Ed. 2020, 59 (28), 11423–11431. https://doi.org/10.1002/anie.2020004...

    work page doi:10.1038/s41467-024-51274-0 2020

  5. [5]

    (8) Liu, D.; Chen, X.; Ren, M.; Ren, Y.; Yaltaye, M

    https://doi.org/10.1038/s41467-019-12690-9. (8) Liu, D.; Chen, X.; Ren, M.; Ren, Y.; Yaltaye, M. W.; Huang, J.-A.; Tian, R.; Liang, L.; Zeng, C. Single-Molecule Resolution of Oligopeptides in Anti-Aging Cosmetics Combined with Nanopore Readouts and Deep Learning Model. Chemical Engineering Journal 2025, 519, 165442. https://doi.org/10.1016/j.cej.2025.1654...

    work page doi:10.1038/s41467-019-12690-9 2025

  6. [6]

    (18) Xin, P.; Zhao, Y.; Liang, Y.; Yaltaye, M

    https://doi.org/10.1021/acsnano.4c09483. (18) Xin, P.; Zhao, Y.; Liang, Y.; Yaltaye, M. W.; Hubarevich, A.; Pankratova, V.; Wang, S.; Huang, J.-A. Single-Molecule Monitoring of Nucleic Acid Dynamics Using Raman Correlation Spectroscopy in Plasmonic Nanowells. ACS Nano 2025, 19 (42), 37241–37254. https://doi.org/10.1021/acsnano.5c12000. (19) Huang, J.-A.; ...

    work page doi:10.1021/acsnano.4c09483 2025

  7. [7]

    (20) Petushkova, N

    https://doi.org/10.1038/s41467-019-13242-x. (20) Petushkova, N. A.; Zgoda, V. G.; Pyatnitskiy, M. A.; Larina, O. V.; Teryaeva, N. B.; Potapov, A. A.; Lisitsa, A. V. Post-Translational Modifications of FDA-Approved Plasma Biomarkers in Glioblastoma Samples. PLoS One 2017, 12 (5), e0177427. https://doi.org/10.1371/journal.pone.0177427. (21) Sugiyama, N.; Ma...

    work page doi:10.1038/s41467-019-13242-x 2017

  8. [8]

    (23) Wang, Z.; Ranasinghe, J

    https://doi.org/10.1038/ncomms9202. (23) Wang, Z.; Ranasinghe, J. C.; Wu, W.; Chan, D. C. Y.; Gomm, A.; Tanzi, R. E.; Zhang, C.; Zhang, N.; Allen, G. I.; Huang, S. Machine Learning Interpretation of Optical Spectroscopy Using Peak-Sensitive Logistic Regression. ACS Nano 2025, 19 (16), 15457–15473. https://doi.org/10.1021/acsnano.4c16037. (24) Wang, Z.; Ye...

    work page doi:10.1038/ncomms9202 2025

  9. [9]

    (27) Shi, J.; Bera, K.; Mukherjee, P.; Alex, A.; Chaney, E

    https://doi.org/10.1021/acsphotonics.5c02676. (27) Shi, J.; Bera, K.; Mukherjee, P.; Alex, A.; Chaney, E. J.; Spencer-Dene, B.; Majer, J.; Marjanovic, M.; Spillman, D. R.; Hood, S. R.; Boppart, S. A. Weakly Supervised Identification and Localization of Drug Fingerprints Based on Label-Free Hyperspectral CARS Microscopy. Anal. Chem. 2023, 95 (29), 10957–10...

    work page doi:10.1021/acsphotonics.5c02676 2023

  10. [10]

    (33) Luz-Lima, C.; De Sousa, G

    https://doi.org/10.1038/s41467-023-42943-7. (33) Luz-Lima, C.; De Sousa, G. P.; Lima, J. A.; Melo, F. E. A.; Filho, J. M.; Polian, A.; Freire, P. T. C. High Pressure Raman Spectra of β-Form of l-Glutamic Acid. Vib. Spectrosc. 2012, 58, 181–187. https://doi.org/10.1016/j.vibspec.2011.12.005. (34) Mary, M. B.; Ramakrishnan, V. Infrared and Laser Raman Spect...

    work page doi:10.1038/s41467-023-42943-7 2012

  11. [11]

    (36) Freire, P

    https://doi.org/10.1016/j.molstruc.2024.138587. (36) Freire, P. T. C.; Barboza, F. M.; Lima, J. A.; Melo, F. E. A.; Filho, J. M. Raman Spectroscopy of Amino Acid Crystals. In Raman Spectroscopy and Applications; InTech,

    work page doi:10.1016/j.molstruc.2024.138587 2024

  12. [12]

    (37) Šapić, M

    https://doi.org/10.5772/65480. (37) Šapić, M. Experimental and Theoretical Study of Aspartic Acid. (38) Jumabaev, A.; Holikulov, U.; Hushvaktov, H.; Absanov, A.; Bulavin, L. INTERACTION OF VALINE WITH WATER MOLECULES: RAMAN AND DFT STUDY. Ukrainian Journal of Physics 2022, 67 (8), 602–610. https://doi.org/10.15407/ujpe67.8.602. (39) Zhu, G.; Zhu, X.; Fan,...

    work page doi:10.5772/65480 2022

  13. [13]

    (51) Wu, Z.; He, H.; Zhao, X.; Lin, Z.; Ye, Y.; Guo, J.; Hu, W.; Jiang, X

    https://doi.org/10.1038/s41467-023-37403-1. (51) Wu, Z.; He, H.; Zhao, X.; Lin, Z.; Ye, Y.; Guo, J.; Hu, W.; Jiang, X. Reimagining Cancer Tissue Classification: A Multi-Scale Framework Based on Multi-Instance Learning for Whole Slide Image Classification. Med. Biol. Eng. Comput. 2025, 63 (9), 2617–2635. https://doi.org/10.1007/s11517-025-03341-x. (52) Che...

    work page doi:10.1038/s41467-023-37403-1 2025

  14. [14]

    A Two-Stage Causality Method for Time Series Prediction Based on Feature Selection and Momentary Conditional Independence

    (53) Ma, D.; Ren, W.; Han, M. A Two-Stage Causality Method for Time Series Prediction Based on Feature Selection and Momentary Conditional Independence. Physica A: Statistical Mechanics and its Applications 2022, 595, 126970. https://doi.org/10.1016/j.physa.2022.126970. (54) Fu, C.; Zhang, K.; Guan, H.; Deng, S.; Sun, Y.; Ding, Y.; Wang, J.; Liu, J. Progr...

    work page doi:10.1016/j.physa.2022.126970 2022

  15. [15]

    (61) Kerker, M.; Siiman, O.; Bumm, L

    https://pubs.acs.org/sharingguidelines. (61) Kerker, M.; Siiman, O.; Bumm, L. A.; Wang, D.-S. Surface Enhanced Raman Scattering (SERS) of Citrate Ion Adsorbed on Colloidal Silver. Appl. Opt. 1980, 19 (19),

    work page 1980

  16. [16]

    (62) Grys, D.-B.; de Nijs, B.; Salmon, A

    https://doi.org/10.1364/AO.19.003253. (62) Grys, D.-B.; de Nijs, B.; Salmon, A. R.; Huang, J.; Wang, W.; Chen, W.-H.; Scherman, O. A.; Baumberg, J. J. Citrate Coordination and Bridging of Gold Nanoparticles: The Role of Gold Adatoms in AuNP Aging. ACS Nano 2020, 14 (7), 8689–8696. https://doi.org/10.1021/acsnano.0c03050. (63) Mary, Y. S.; Ushakumari, L.; ...

    work page doi:10.1364/ao.19.003253 2020

  17. [17]

    J.; Aliaga, A

    (64) Cárcamo, J. J.; Aliaga, A. E.; Clavijo, E.; Garrido, C.; Gómez-Jeria, J. S.; Campos-Vallette, M. M. Proline and Hydroxyproline Deposited on Silver Nanoparticles. A Raman, SERS and Theoretical Study. Journal of Raman Spectroscopy 2012, 43 (6), 750–755. https://doi.org/10.1002/jrs.3092. (65) Jarmelo, S.; Reva, I.; Carey, P. R.; Fausto, R. Infrared and ...

    work page doi:10.1002/jrs.3092 2012

  18. [18]

    https://doi.org/10.1002/jrs.954

    Journal of Raman Spectroscopy 2003, 34 (1), 77–83. https://doi.org/10.1002/jrs.954. (71) Kolesov, B. A.; Boldyreva, E. V. Difference in the Dynamic Properties of Chiral and Racemic Crystals of Serine Studied by Raman Spectroscopy at 3−295 K. J. Phys. Chem. B 2007, 111 (51), 14387–14397. https://doi.org/10.1021/jp076083o. (72) Jarvis, R. M.; Blanch, E. W.;...

    work page doi:10.1002/jrs.954 2003

  19. [19]

    https://doi.org/10.1039/b702944f

    work page doi:10.1039/b702944f