**1. Introduction**

Amyloidosis is an uncommon systemic disease caused by irregular protein aggregation and misfolding that leads to the formation of insoluble amyloid deposits [1–4]. Different types of amyloid derive from various amyloid precursor proteins and can infiltrate various organs [1,5]. Although these protein deposits and their sequences vary, amyloid fibrils share a common structure, namely steric zippers, arranged in a periodic fibrillar lattice of β-sheets; this structure can be observed across various modalities, including NMR spectroscopy, cryo-electron microscopy (cryo-EM), and atomic force microscopy (AFM) [6–8].

Recently, Raman spectroscopy has been utilized to study amyloid fibril formation and structural conformations [9–13]. By vibrationally fingerprinting biological samples at a molecular level, Raman spectroscopy identifies various molecules, including proteins and lipids, with high sensitivity and in a nondestructive and label-free manner [14–20]. In addition, its relatively simple setup and the lack of a requirement for a priori knowledge of sample composition make Raman spectroscopy a potential tool to study amyloidosis. Previous studies have shown that Raman spectroscopy is sensitive to differences in structural conformations of different amyloid types [11,12,21–23]. In particular, amide I and

**Citation:** Kim, J.H.; Zhang, C.; Sperati, C.J.; Bagnasco, S.M.; Barman, I. Non-Perturbative Identification and Subtyping of Amyloidosis in Human Kidney Tissue with Raman Spectroscopy and Machine Learning. *Biosensors* **2023**, *13*, 466. https:// doi.org/10.3390/bios13040466

Received: 16 March 2023 Revised: 5 April 2023 Accepted: 6 April 2023 Published: 8 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

III bands identified β-sheet structures in both amyloid fibrils isolated from patients and synthesized amyloid peptides [9,10,22,24]. However, although these findings established the applicability of Raman spectroscopy to study amyloidosis, synthesized amyloid and isolated amyloid fibrils are overly simplified and disconnected from protocols of clinical detection and diagnosis.

To address this limitation, several researchers have investigated amyloid deposits in tissue with Raman spectroscopy. Animal models have been used to identify biomarkers representative of the amyloid signature within a mixture of biomolecules, coupled with spectral unmixing analysis [25–27]. In addition, others have applied Raman spectroscopy to tissue biopsies of patients that reported changes in the protein signature associated with amyloid [24,28–33]. Although these studies demonstrate Raman spectroscopy's capability to distinguish subtle spectral changes due to amyloid deposits in tissue samples, they were mainly focused on brain tissues to investigate amyloid involvement with disorders such as Alzheimer's disease and Parkinson's disease. However, no previous Raman spectroscopic investigations of renal amyloid deposits exist, despite the fact that the kidney is one of the most commonly involved organs in amyloidoses [5,34].

Here, we employ Raman spectroscopy to examine amyloid deposits for the first time, to the best of our knowledge, in unstained fresh-frozen human kidney tissues. Specifically, we investigated immunoglobulin light chain (AL) and serum amyloid A (AA), which are precursor proteins that give rise to AL amyloidosis and AA amyloidosis, respectively [3,35]. These amyloid diseases represent the two major amyloid diseases with kidney involvement [5,36,37]. We investigated the Raman spectra of AL, AA, and non-amyloidogenic (NA) tissues collected from six patients through analyses of the protein band area and second derivative. Then, using t-distributed stochastic neighbor embedding (t-SNE) and density-based spatial clustering of applications with noise (DBSCAN), we characterized endogenous molecular compositions and structures indicative of amyloid deposits and demonstrated heterogeneity between different amyloid types. In this study, we describe in detail our methodological approach, combining Raman spectroscopy with machine learning techniques to identify and characterize the two major types of amyloidosis in human renal tissue.
