Exploratory feature analysis of baseline ionic currents in MoS₂ nanopores

Ionic-current fluctuations in nanometer-scale pores are relevant to hydrophobic gating in neuromorphic memristive systems and to signal interpretation in nanopore biosensing. While nanopore measurements commonly focus on transient current blockades, the statistical structure of baseline currents recorded in the absence of analytes has received less attention. Here, baseline recordings from two-dimensional MoS₂ nanopores were analyzed to identify recurring signal descriptors associated with operationally assigned electrical regimes. Constant-voltage recordings were partitioned into analysis windows, from which statistical, temporal, and spectral descriptors were extracted and represented using kernel principal component analysis (Kernel PCA). Density-based clustering was used to identify outlying samples, while procrustes analysis ranked feature combinations according to how closely they preserved the geometry of the resulting embedding. Spectral descriptors were recurrent among the top-ranked features of the open-associated reference dataset, whereas the closed-associated reference dataset showed stronger contributions from temporal variability together with additional spectral descriptors. The mixed/transition-associated dataset exhibited combinations of descriptors observed across both reference regimes. A separate voltage-matched analysis of sequential closed-associated and open-associated sweeps from the same nanopore was included as a within-pore sensitivity analysis. This comparison held pore identity, nominal diameter, KCl concentration, and the applied-voltage set constant, although the sequential acquisition design does not exclude time- or order-dependent effects. Overall, baseline ionic-current recordings exhibited structured multivariate signatures that could be characterized through geometry-preserving feature analysis. This work proposes an exploratory analysis of the nanopore electrical-regime-associated signals without the need of direct data-intensive classification schemes.