# One-Drop Serum Screening Test to Monitor Tissue Iron Accumulation

## TL;DR

This study introduces a minimally invasive blood test using infrared spectroscopy and machine learning to detect and measure iron overload in blood and tissues without biopsies.

## Contribution

A novel one-drop serum screening method for identifying iron overload and quantifying iron levels in multiple tissues using Fourier transform infrared spectroscopy and machine learning.

## Key findings

- PLS-DA and PLS regression models accurately classified and quantified iron levels in blood and tissues.
- Spectral analysis revealed functional interrelationships between spleen-liver and heart-kidney pairs.
- The method demonstrated excellent linearity and low error in quantifying iron concentrations.

## Abstract

Although iron is an essential element for vital body
functions,
iron overload (IO) is accompanied by significant cellular damage due
to its accumulation within organs. Thus, early diagnosis and accurate
identification of the affected organs are critical for preventing
irreversible damage and improving patient survival rates. Diagnosing
tissue iron deposits relieves invasive biopsies with atomic absorption
spectrometry (reserved for specific cases) or noninvasive but costly
and time-consuming imaging techniques like computerized tomography
and magnetic resonance, which provide limited analytical data and
are unsuitable for routine screening. As an alternative, Fourier transform
infrared spectroscopy combined with machine learning has emerged as
a promising approach for supporting medical decision-making. In this
study, we developed a minimally invasive method to identify IO and
quantify iron levels in blood and tissues (heart, liver, spleen, and
kidney) without biopsies. PLS-DA classification models and PLS regression
models were constructed based on samples categorized into a control
group (n = 10) and three iron-administered groups
at 250 mg kg–1 (n = 14), 500 mg·kg–1 (n = 13), and 1000 mg·kg–1 (n = 15). Iron levels were measured
in blood samples and tissue biopsies (spleen, heart, liver, and kidney).
The binary classification models (control vs iron-administered) and
multiclass models (control, 250, 500, and 1000 mg·kg–1) demonstrated satisfactory performance into train and validation
groups. PLS regression models for quantifying iron concentrations
in blood and tissues exhibited excellent linearity and low associated
errors across both calibration and test groups. Permutation tests
confirmed that all models found a real class structure in the data,
were not random, and were built using true chemical information. The
chemical insights from the spectra may reflect adaptations associated
with iron-induced dysregulation. Alterations in biomolecules could
reflect systemic stress responses and may result from free radicals
generated by the iron-induced Fenton reaction. Moreover, key spectral
regions revealed functional interrelationships, particularly between
spleen and liver, and heart and kidneys. In summary, the findings
support the potential of this innovative for future research to identify
IO and quantify iron levels in human blood and different human tissues
using only a single drop of blood without tissue biopsies.