A Comparison between Different Machine Learning Approaches Combined with Anodic Stripping Voltammetry for Copper Ions and pH Detection in Cell Culture Media

Abstract

Recently, the scientific community has shown a great interest about the Organ-on-Chip (OoC) devices, a special kind of micro-fabricated platforms capable of recapitulating the human physiology implementing the traditional cell culture methods and the concept of in vivo studies. Copper ions represent a cellular micronutrient that must be monitored for its potential hazardous effects. The application of electrochemical analysis for heavy metal ions detection and quantification in commercial cell culture media presents several issues due to electrolyte complexity and interferents. In fact, to the best of our knowledge, there is a lack of applications and OoC devices that implement the Anodic Stripping Voltammetry as an ion dosing technique due to the reasons reported above. In fact, considering just the peak intensity value from the measurement, it turns out to be challenging to quantify ion concentration since other ions or molecules in the media may interfere with the measurement. With the aim to overcome these issues, the present work aims to develop an automated system based on machine learning algorithms and demonstrate the possibility to build a reliable forecasting model for copper ion concentration on three different commercial cell culture media (MEM, DMEM, F12). Effectively, combining electrochemical measurements with a multivariate machine learning algorithm leads to a higher classification accuracy. Two different pH media conditions, i.e., physiological (pH 7.4) and acidic (pH 4), were considered to establish how the electrolyte influences the measurement. The experimental datasets were obtained using square-wave anodic stripping voltammetry (SWASV) and were used to carry out a machine learning trained model. The proposed method led to a significant improvement in Cu2+ concentration detection accuracy (96.6% for the SVM model and 93.1% for the NB model in MEM) as well as being able to monitor the pH solution.

Keywords

• chemometrics
• machine learning
• classification
• heavy metal ion detection
• complex media