Quantification of L-lactic acid in human plasma samples using Ni-based electrodes and machine learning approach

Area of Science:

• Electrochemistry
• Biosensors
• Machine Learning

Background:

• Electrochemical sensors often face challenges with selectivity and interference from complex biological samples.
• Accurate quantification of analytes like L-lactic acid in human plasma is crucial for diagnosing metabolic conditions.

Purpose of the Study:

• To develop a robust strategy for quantifying overlapping electrochemical signatures in complex mixtures and human plasma.
• To create a non-enzymatic electrochemical sensor for L-lactic acid detection using nickel oxide nanoparticles.
• To utilize machine learning models to manage interference and enhance sensor performance.

Main Methods:

• Fabrication of nickel oxide (NiO) nanoparticle-modified glassy carbon electrodes (GCE) for electrochemical sensing.
• Measurement of interference trends from uric acid, ascorbic acid, and glucose in the presence of L-lactic acid.
• Analysis of a large dataset using various machine learning models, including artificial neural networks and random forests.
• Validation of sensor performance using 10 different human plasma samples and comparison with conventional colorimetric assays.

Main Results:

• Achieved limits of detection (LOD) of 2.61 μM for L-lactic acid, 15.99 μM for uric acid, 11.34 μM for glucose, and 3.27 μM for ascorbic acid in complex mixtures using an artificial neural network.
• Demonstrated good prediction performance in human plasma samples with a random forest model, yielding R²=0.99, LOD=1.3 μM, and LOQ=4.4 μM for L-lactic acid.
• Successfully validated the sensor's performance in real human plasma samples, showcasing its potential for clinical applications.

Conclusions:

• The developed strategy effectively quantifies overlapping electrochemical signatures, overcoming transducer selectivity limitations through machine learning.
• The nickel-based electrochemical sensor coupled with ML models shows significant promise for analyzing metabolic profiles in complex biological samples.
• Miniaturization and integration into point-of-care testing devices could enable real-time monitoring of metabolic conditions, such as sepsis.