Noninvasive inference of presymptomatic abiotic stress in Solanum Lycopersicum from latent electrophysiological dynamics via action potential-based model

is critical for yield protection, yet current methods rely on visible symptoms that manifest only after irreversible physiological damage. Plant action potentials (APs), which directly reflect ion channel activity across the plasma membrane, represent an underexplored avenue for presymptomatic, noninvasive stress detection. We hypothesized that stress-induced alterations in ion channel activity would produce measurable changes in AP waveform morphology before visible symptoms appeared. One hundred sixty S. lycopersicum cv. Moneymaker plants, spanning two prospectively defined cohorts (80 for classifier development, 80 for detection-window quantification), were assigned to control, salinity (100 mM NaCl), osmotic (PEG-6000, -0.4 MPa), or K⁺/N deficiency groups. APs were recorded noninvasively using a Plant SpikerBox, and five physiologically interpretable parameters were extracted via phenomenological curve fitting. Two models were evaluated: logistic regression (LR) on biophysical parameters and a convolutional neural network (CNN) on raw waveforms. Assessed on a single locked held-out partition of 39 recording sessions shared by both models, LR did not underperform the CNN (accuracy 0.923 vs. 0.897; ROC-AUC 0.955 vs. 0.976). The paired difference in ROC-AUC was not distinguishable from zero (ΔAUC = -0.022, 95% CI -0.106 to +0.041), and confidence intervals are wide and overlapping at this sample size; the comparison should be read as a failure to detect a difference rather than as a demonstration of equivalence. Applied prospectively to an independent cohort, the classifier detected stress 5.25 ± 2.96 days before visible symptom onset across all three stress conditions, with all individual presymptomatic detection windows positive, while issuing no sustained positive call on any concurrently recorded control plant. This work establishes individual tomato plant action potentials as quantitative probes of physiological state, with direct implications for precision agriculture.

Authors

Publication Details

Journal
Plant Signaling & Behavior
Published
2026-11-16
DOI
https://doi.org/10.1080/15592324.2026.2750612
Primary Topic
Plant and Biological Electrophysiology Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Noninvasive inference of presymptomatic abiotic stress in Solanum Lycopersicum from latent electrophysiological dynamics via action potential-based model

권율희, Jacob Wu, Rudra Pachori
1 links
Plant Signaling & Behavior
Plant and Biological Electrophysiology Studies
article

Noninvasive inference of presymptomatic abiotic stress in Solanum Lycopersicum from latent electrophysiological dynamics via action potential-based model

권율희, Jacob Wu, Rudra Pachori
article en

Abstract

is critical for yield protection, yet current methods rely on visible symptoms that manifest only after irreversible physiological damage. Plant action potentials (APs), which directly reflect ion channel activity across the plasma membrane, represent an underexplored avenue for presymptomatic, noninvasive stress detection. We hypothesized that stress-induced alterations in ion channel activity would produce measurable changes in AP waveform morphology before visible symptoms appeared. One hundred sixty S. lycopersicum cv. Moneymaker plants, spanning two prospectively defined cohorts (80 for classifier development, 80 for detection-window quantification), were assigned to control, salinity (100 mM NaCl), osmotic (PEG-6000, -0.4 MPa), or K⁺/N deficiency groups. APs were recorded noninvasively using a Plant SpikerBox, and five physiologically interpretable parameters were extracted via phenomenological curve fitting. Two models were evaluated: logistic regression (LR) on biophysical parameters and a convolutional neural network (CNN) on raw waveforms. Assessed on a single locked held-out partition of 39 recording sessions shared by both models, LR did not underperform the CNN (accuracy 0.923 vs. 0.897; ROC-AUC 0.955 vs. 0.976). The paired difference in ROC-AUC was not distinguishable from zero (ΔAUC = -0.022, 95% CI -0.106 to +0.041), and confidence intervals are wide and overlapping at this sample size; the comparison should be read as a failure to detect a difference rather than as a demonstration of equivalence. Applied prospectively to an independent cohort, the classifier detected stress 5.25 ± 2.96 days before visible symptom onset across all three stress conditions, with all individual presymptomatic detection windows positive, while issuing no sustained positive call on any concurrently recorded control plant. This work establishes individual tomato plant action potentials as quantitative probes of physiological state, with direct implications for precision agriculture.

Plant Signaling & BehaviorVol. 21(1)
Zero hunger
Plant and Biological Electrophysiology Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon