Microfluidics-based lab-on-a-chip systems, which feature miniaturization of biological separation and assay techniques, are rapidly transforming biochemical analysis and high-throughput screening. Microfluidic system design requires expertise in materials, chemistry, biology, and engineering, and understanding the complex interplay between variables that influence and limit system performance is difficult without computational assistance. Modeling approaches based on 3-dimensional numerical simulations provide detailed information regarding spatiotemporal variations of the field variables but are computationally very expensive for system-level analysis. Design-modeling tools are needed that rapidly simulate the complex underlying phenomena such as electroosmosis, electrophoresis, sample dispersion, mixing, and biochemical reactions without significantly compromising accuracy (1)(2)(3)(4)(5). In addition, these design tools must be easily usable by the microfluidic community, which comprises scientists and engineers from a variety of disciplines. To meet these challenges, we developed integrated design software that allows rapid layout of microfluidic channel networks, fast system performance simulation using a system solver, and the ability to easily reconfigure chip layout to meet specifications. We illustrate the application of the software to improve the design for an electrokinetic immunoassay chip.

The software design follows a modified form of the traditional client-server architecture. The user interacts with a graphical user interface (GUI) front-end (client). Fig. 1 shows a screenshot of the GUI. The microfluidic lab-on-a-chip system is represented as a network of interconnected components that can be assembled from a component library. The sequence of operations required for the creation of the microfluidic network, analysis, and visualization of results using the GUI is as follows:

(1) Creation of the microfluidic network: The components (such as sample reservoirs, straight channels, bends, biosensors, and interconnects such as Y-, T-, and cross junctions) are selected from the component library and assembled into a network using a drag-and-drop method.

(2) Problem specification: Geometric properties for components (channel length, breadth, and depth, turn radius and angle, well diameter) and operating conditions (applied voltage and pressure, flow rate, and injected analyte concentrations, as appropriate for the problem under consideration) are specified for the components. The property database contains physicochemical property data for commonly used buffers, reagents, and analytes (density and viscosity of buffers; electrical conductivity, and electrokinetic mobility and molecular diffusivity of analytes) and is fully integrated with the GUI.

(3) Solution and visualization: The performance is simulated using the system solver, and the results are analyzed using the visualization toolkit, which allows the results from the simulation to be displayed in a variety of tabular and graphical formats.

The GUI employs the hierarchical model-view-controller (MVC) architecture (6) to achieve the user-friendliness, flexibility, and extensibility needed. MVC programming uses 3-way factoring, whereby objects of different software classes take over the operations related to the application domain (model), the display of the application’s state (view), and the user interaction with the model and the view (controller). The MVC architecture is aimed at exploiting the benefits associated with modular components in the software. The GUI has been developed with the Java™ programming language using standard Java libraries and the included Swing toolkit(7).

(4) System solver: The system solver uses a combination of various modeling approaches for a rapid simulation of the microfluidic chip performance. This mixed-methodology approach uses an integral method to simulate fluid flow and electric field, a method of moments-based analytical solution to compute analyte dispersion, and a Fourier series–based analytical solution to compute microfluidic mixing based on laminar diffusion. These disparate models have been integrated in the system solver and validated against both experimental data and detailed 3-dimensional numerical models. The system solver shows a substantial improvement in computational speed (2–4 orders of magnitude) over the 3-dimensional models without appreciably compromising accuracy (error <10%). Details of the models and validation studies have been described elsewhere (4)(8)(9)(10). A brief explanation of these models is given below:

  • Fluid flow: Pressure-driven flow is calculated by solving the Navier-Stokes and continuity conservation equations in their integral forms. An implicit iterative numerical solution scheme based on the SIMPLE (semiimplicit method for pressure-linked equations) algorithm (11) is used. Details of the implementation are discussed elsewhere(8).

  • Electric field: The electric current conservation law is solved at every component with a constitutive equation to compute currents and voltages. These equations are used to compute the electroosmotic and electrophoretic flow velocity.

  • Analyte transport: An analytical model based on the method of moments approach has been developed to characterize the dispersion induced by combined pressure and electrokinetic-driven flow. In addition, the system solver uses a combination of numerical schemes and analytical approaches to simulate mixing due to laminar diffusion and biochemical reactions; specifically, the method of lines (MOL) and 2-compartment models for biochemical reactions, and a Fourier series-based model for analyte mixing.

We present the use of the microfluidic design software to improve the design of an electrokinetic microfluidic device for an on-chip assay of the drug theophylline (Th) in serum samples. The assay involves on-chip mixing of serum samples with a labeled tracer compound and reaction with a selective antibody. This reaction is followed by an electrophoresis-based separation step to isolate and quantify the reactants and products. This immunoassay method is appropriate for incorporation into a microfluidic format and allows for rapid separation, because of the short separation distances. Starting with the microfluidic device previously demonstrated (12), we applied the software to rapidly explore alternative design concepts to improve device performance and demonstrate the ability to create a lab-on-a-chip for a real clinical analysis using a simulation-based design approach. Similar analysis using an analog hardware description language (Verilog-A) has been previously reported(13).

The operation of the competitive immunoassay chip is a multistep process that includes (a) mixing of serum sample containing Th with fluorescein-labeled Th tracer (Th*), carried out in the microfluidic channel and based on laminar diffusion; (b) reaction of the resulting mixture with an anti-Th antibody (Ab), which allows Th and Th* to compete for a limited number of antibody-binding sites; (c) electrokinetic injection of the solution containing the Ab-Th* complex produced in the reaction, as well as the unreacted Th*, into the separation channel, where they are separated by electrophoresis; and (d) detection of the fluorescent species (Th* and Ab-Th*) by laser-induced fluorescence.

The chip layout was created in the layout editor, using the component library and the drag-and-drop methodology. The layout parameters (channel dimensions, connectivity) and operating conditions (voltages and concentrations) were specified, and the performance of the chip was simulated. All channels had a rectangular cross-section with a uniform depth of 20 microns. The original layout had channels with widths of 52–236 microns, and the same widths were used in the modified layout. Th and Th* in the sample were specified in the system design software as separate species with identical molecular diffusivity (3.3 × 10−10 m2/s) and electrokinetic mobility [2.84 × 10−8 m2/(V s)]. The corresponding properties for the antibody (Ab) were 4.0 × 10−11 m2/s (molecular diffusivity) and 4.45 × 10−8 m2/(V s) (electrokinetic mobility) (14). The binding between Th and Ab was assumed to be irreversible and complete. An electric field of 770 V/cm was used for electrophoretic separation. The performance was characterized by the extent of mixing/reaction and by the efficiency of separation, which is characterized by the separation resolution, peak height, variance, and time for separation. The layout was reconfigured using the layout editor to minimize the chip footprint. The modified layout (3.5 cm × 3.5 cm) occupies <25% of the area of the original chip (7.6 cm × 7.6 cm), and the time required for electrophoretic separation was decreased by more than 50% relative to the original design(12), thereby decreasing the overall assay time. This reconfiguration decreased the separation resolution, but the resulting resolution was still sufficient to resolve the species bands while limiting the band-broadening induced by dispersion. In addition, the signal amplitude increased by 11.5% and 5% for Ab-Th* and Th*, respectively. The degree of mixing for the antibody: Formula

where y is the widthwise coordinate, w is the channel width, c is the concentration profile along the channel width, and cavg is the average concentration along the width, was also improved to 100%. The entire analysis (including layout generation and problem setup) was completed in approximately 4 h, more than 2 orders of magnitude faster than currently available techniques. The improvements are summarized in Table 1 . In retrospect, the original system was substantially overdesigned, a problem that is common to several microfluidic systems currently available today and is attributable primarily to a lack of design tools.

In summary, the design software we describe is useful for estimating device performance and creating microfluidic chip layouts; these layouts can be rapidly modified to design chips that meet performance requirements. We used the software to improve the design of a microfluidic immunoassay chip. The resulting design occupied <25% of the area of the original chip, and the time required for electrophoretic separation was decreased by more than 50% relative to the original design, allowing for a faster assay. This process is more than 2 orders of magnitude faster than conventional design techniques and is ideally suited for design optimization of microfluidic lab-on-a-chip systems.

SOURCE:  http://www.clinchem.org/content/53/11/2023.full