Multisensor microfluidic chip and machine learning for the pharmacokinetic analysis of patient breast cancer cell viability: Enhancing accessibility and scale of in vitro antineoplastic testing in precision oncology

• Grant number: 25/00614-6
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: June 01, 2025
• End date: July 31, 2028
• Field of knowledge: Physical Sciences and Mathematics - Chemistry - Analytical Chemistry
• Principal Investigator: Renato Sousa Lima
• Grantee: Bruna Gabrielle Olsen Pinto
• Host Institution: Centro Nacional de Pesquisa em Energia e Materiais (CNPEM). Ministério da Ciência, Tecnologia e Inovação (Brasil). Campinas , SP, Brazil
• Associated research grant: 23/00246-1 - Miniaturized large-scale devices for in-situ analysis: fabrication, characterization and applications, AP.TEM

Abstract

In vitro tests of cell susceptibility to chemotherapy are important for precision oncology, as they support clinical decisions regarding the safest and most effective chemotherapy strategy in a precise and personalized manner, contributing to increasing patient survival. In this case, the in vitro assessment of the effect of drugs is done by analyzing the viability (number of live cells) of primary cell models (obtained from the patient), 2D or 3D. However, traditional methods of analyzing this parameter are often expensive, time-consuming and invasive. A faster study of the effects of drugs increases assertiveness and reduces the time and costs required to define the most effective treatment for patients. On the other hand, non-invasive analyses allow real-time tests (pharmacokinetics) that help us study the mechanism of action of drugs. Therefore, in this doctoral project, we propose an accessible method for performing in vitro pharmacokinetic tests of the susceptibility of primary breast cancer cells, the most common type among women, to various commercial chemotherapeutics with high throughput and accuracy. To ensure low-cost attributes, real-time measurements and throughput, microfluidic multisensor electrochemical chips will be used for electrochemical tests of 2D cell detachment from patients with that type of tumor via non-invasive, rapid, serial analyses and at different times (of exposure to the drug). The chip will operate as a multifunctional platform by enabling (i) the proliferation of 2D cells and then (ii) the determination of cell viability. This determination will be mediated by machine learning (ML), which will generate a model for predicting cell viability from the sensor data and, then, from the half-median concentration (LC50) of the drugs, another crucial pharmacological data. This model should be universal, regardless of the drug tested. The analyses will be performed on a glass chip with a large number of sensors, which are arranged vertically in a grid pattern. Developed in our group, this sensor arrangement is integrated into a compact chip. Therefore, the sensors combine high resolution and reproducibility (merits of microfabrication) with low cost due to the large number of sensors per area. Furthermore, due to its compact size, the chip is compatible with microfluidics, a tool that reduces reagent consumption and increases reproducibility. In this project, microfluidic chips will be obtained by sealing channels in polydimethylsiloxane substrates. The devices will have 9 channels, with 5 sensors in each of them (total of 45 sensors). Therefore, up to 9 samples can be analyzed on a single chip, with independent measurements in five times for each case. The electrochemical cell detachment tests (cells detach from the electrode surface after their drug-induced death) will be based on the variation of the steric hindrance imposed on the redox probe Ru(NH3)63+, which will be added to the medium and monitored by square wave voltammetry (SWV). Our strategy to increase the testing capacity will be based on (i) the use of multisensor chips to accommodate several samples simultaneously, (ii) performing individual fast (3 s) SWV electrochemical tests, and (iii) performing serial analyses using a portable single-channel potentiostat. Since each SWV measurement will last 3 s, the analyses of all 45 sensors on the chip will be completed in only 2 min and 15 s. The SWV technique will be used in order to obtain not only fast analyses, but also multivariate data that will be processed by ML for the accurate determination of cell viability.