ML-automated microfluidic circuit design.

Area of Science:

• Biotechnology
• Engineering
• Computer Science

Background:

• Microfluidic chip design demands specialized expertise and iterative processes, limiting accessibility for non-specialists.
• Current methods for microfluidic fabrication present significant barriers to entry for researchers without extensive experience.

Purpose of the Study:

• To introduce μFluidicGenius (μFG), an open-access, machine learning-augmented design tool for rapid microfluidic circuit creation by nonexperts.
• To enable users to define microfluidic layouts, including reservoir placement, channel connections, and flow rates, for automated design generation.

Main Methods:

• Utilized a hybrid algorithmic framework combining machine learning (ML) models and mathematical modeling.
• Developed a system that generates spatially coded maze structures to achieve precise fluidic resistances for target flow distributions.
• Designs are optimized for geometry and exportable for 3D printing.

Main Results:

• μFG successfully generates microfluidic designs that implement precise fluidic resistances to meet specified flow rates.
• The tool can reproduce complex flow profiles, including those relevant for multi-organ-on-chip applications.
• Experimental validation confirmed that μFG-generated circuits achieve 90% accuracy in reproducing target flow distributions.

Conclusions:

• μFluidicGenius (μFG) significantly lowers the barrier to entry for microfluidic chip design, empowering nonexperts.
• Demonstrates an effective application of ML in automating and streamlining the design of complex microfluidic architectures.
• Facilitates rapid, customizable, and accurate development of microfluidic systems for diverse applications.