Hardware

Overview

In the design process of our colloidal gold test strips, we plan to utilize de novo protein design to obtain protein binders for GZMK. However, we've found that the success rate of de novo protein design is low. It often requires extensive expression and screening to obtain proteins with high affinity. Traditional expression systems are inefficient and costly, leading to a significant investment of time and resources to acquire protein binders with ideal properties.

To address this challenge, we have developed Nexus, a high-throughput affinity measurement platform. This platform integrates a cell-free expression system with affinity detection, enabling the rapid synthesis and initial affinity screening of protein binders. For the high-affinity candidate proteins, we then use high-precision instruments for precise measurement, efficiently obtaining protein binders with the desired functions

Our hardware and wet lab teams have collaborated closely throughout the platform's development, continuously iterating based on practical needs. After multiple rounds of sample testing and design optimization, we built the high-throughput affinity detection platform into an integrated architecture, which makes the input and output of samples easier to control, improves the efficiency of affinity measurement, and enables the testing and analysis results to be obtained in a simpler manner.

Integrated Architecture

Our integrated architecture is composed of three main components: Microfluidic Chips, Automated Control Equipment, and the Software Platform.

The microfluidic chips enable high-throughput affinity detection of microscale protein samples.

The automated control equipment achieves automatic sample input and output, ensuring precise and efficient fluid manipulation.

The software platform integrates device control, experimental workflow management, data analysis, and visualization, forming a closed-loop process.

Microfluidic Chips

Background

We selected the cell-free protein expression system to express de novo designed protein binders. However, under traditional laboratory conditions, such systems typically have small expression volumes and high cost per unit yield. The expression volume of a single system is nearly equivalent to the sample volume required for one measurement in a microplate reader, which greatly limits the throughput of experiments and the reproducibility of subsequent detections.

Meanwhile, we chose Fluorescence Cross-Correlation Spectroscopy (FCCS) as the method for affinity detection. This technique requires minimal sample consumption and offers high temporal resolution, allowing real-time monitoring of molecular interactions in solution systems.

Based on these considerations, we developed a microfluidic chip–based detection platform that significantly reduces the sample volume required for a single measurement. The system is designed with fully transparent materials to facilitate FCCS detection.

Engineering

Besides the construction of the detection platform, we also optimized both the optical performance and the mixing efficiency.

(click the link to learn more)

Result

After several iterations, our final chips adopt a "primary-secondary collaborative" architecture. The secondary chip is responsible for generating a concentration gradient, while the primary chip handles mixing and FCCS detection.

Both the primary chip and secondary chip are fabricated by bonding PDMS to 0.15 mm-thick, and are assembled in a stacked configuration, forming an interconnected system.

Secondary Chip

The secondary chip serves as the core of gradient formation, the N-to-1 miniature chip with a ballast incorporates an improved tubing connection scheme, delivering higher connection stability and reliability and supporting repeated plug-and-play without damage. The secondary chip enables pre-dilution and facilitates subsequent automation integration; it can be expanded with different auxiliary modules according to experimental needs, enhancing stability and scalability.

Using a 2-to-1 chip as an example, the operational stability of the secondary chip was evaluated.

The results show that this primary–secondary chip strategy mixes and dilutes different liquids, such as cell-free systems and protein solutions, effectively.

Primary Chip

Protein solutions with gradient concentrations(Liquid A) output from the secondary chip were introduced into the input channel of the primary chip, while the binding partner protein(Liquid B) at a fixed concentration was injected from another inlet. The two solutions mixed within the reaction channels. When the mixed solution flowed to the detection window (red circle), FCCS detection was performed.

In the experimental validation, a 488 nm standard and a 561 nm standard were separately introduced into the primary chip, and the sample was subjected to FCCS detection in the detection window. Although there was noise in the 561 nm channel and the cross-correlation curve, the overall curves were within the normal range. From the analysis, binding data between the two proteins could be obtained.

3D-Printed Chip Cartridge

Furthermore, we need to integrate one primary chip and one secondary chip into a chip cartridge to consolidate the functions of input, dilution, mixing, detection, and output within the smallest possible system. This is to facilitate portability and ease of use. We need to design a housing for the primary and secondary chips to achieve unified packaging. The primary chip's dimensions are 75×25mm, the secondary chip's PDMS layer is 25×30mm with a thickness of 3-5mm, the channel depth and width are 50×100μm, the substrate glass slide is 75×25×1mm, and the punched hole size is 1.4mm.

The cartridge must be designed to integrate the sample injection function of the secondary chip with the protein mixing function of the primary chip. The specific solution involves connecting a thin tube to the inlet port on top of the secondary chip, and another tube from its outlet port to the sample inlet of the primary chip. The proteins will then mix in the primary chip and exit through its outlet, thereby achieving high-throughput sample injection. Furthermore, since different secondary chips are used to inject various proteins, it must be possible to easily swap them out during the input process. Structurally, the device needs to provide a reliable clamping mechanism that secures the chips without compressing the PDMS, and it must incorporate a basic leak-proof and visualizable design.

The selection of 3D printing materials is differentiated based on the function of each component. The main housing is made of PLA material, which offers the advantages of minimal warping, dimensional stability, and low cost. The transparent lid is made from a plexiglass sheet, which provides a certain level of mechanical strength while also offering the benefit of transparency for visualization. Stainless steel M4 screws are used as fasteners, and the thin tubing is made of medical-grade silicone.

Automated Control Equipment

In traditional protein affinity detection, sample injection is a manual process. When the scale of microfluidic screening expands to high-throughput, parallel levels, it becomes difficult for manual operation to precisely perform the dilution of multiple concentration gradients. Furthermore, it is impossible to strictly control the timing of when samples of different concentrations enter the detection port. This not only consumes significant labor costs but is also prone to poor experimental reproducibility and reduced data reliability due to human error. To address this, we employ programmable and automated syringe pumps and peristaltic pumps to replace manual operations, thereby automating concentration gradient generation and sample injection timing. This lays the foundation for the high-throughput, high-precision operation of the integrated system.

Based on laboratory conditions and the need for cost reduction, we ultimately adopted a solution of "using a syringe pump to deliver the protein solution and a peristaltic pump to deliver the buffer solution."

For delivering the protein solution, we selected the YHPLC0100S syringe pump, which comes with the pump body, a power adapter, and a controller. This device features low shear force, which effectively prevents protein denaturation, and also possesses high precision, ensuring the accuracy of the reaction system's concentration, thereby guaranteeing the reliability of the core experimental data. The pump uses a stepper motor and a lead screw drive to convert the motor's rotation into a linear push, achieving a stable and quantitative output of liquids or gases. The syringe size is also interchangeable. The compact, independent controller allows for setting speed, operating time, and working mode with just 8 keys. Parameters can be saved after power-off, and multiple interfaces are reserved for easy connection to peripherals. It is small and compact, making it easy to hold or mount. Its stable flow and high precision make it suitable for micro-volume liquid handling and processing. It can also be controlled via communication with dedicated PC software.

For buffer solution delivery, a Lange BT100-2J peristaltic pump was selected. It offers advantages such as multiple channels, low cost, and easy maintenance, making it suitable for delivering the buffer solution, which has lower requirements regarding denaturation and precision. It can meet the needs of high-throughput experiments. Its single-channel flow rate ranges from 0.002 to 380 mL/min, and it supports various pump head configurations as well as control of start/stop, direction, speed, and RS485 communication. The rotation speed can be adjusted manually or via external control.

Automated Measurement under Sequential Control

By using an external syringe pump connected to the chip to control the flow rate, samples are guided to flow slowly and sequentially past a small, transparent detection window on the chip, which is aligned with a fixed detector. The detector remains stationary, and as each concentration sample reaches the detection point, an FCCS signal is collected. This process is repeated to acquire data for all gradient points. The temporal control for measuring affinity using FCCS combined with microfluidics requires a comprehensive consideration of both the detection time for a single concentration point and the interval time for sample switching.

Detection Time for a Single Concentration Point: A signal needs to be collected for 3 to 5 minutes for each concentration point. FCCS requires a sufficient number of photons to calculate an accurate correlation function. Typically, 3 to 5 sets of signals are collected for each point, with each set lasting 30 to 60 seconds. At lower concentrations, the signal is weaker, which may necessitate extending the collection time to 5 minutes.

Interval Time for Sample Switching: A 1 to 2-minute period for rinsing and stabilization is required between two concentration points. After the previous measurement is complete, the microfluidic channel is rinsed with a buffer solution to wash away any residual sample from the previous concentration. The new concentration sample is then allowed to fill the detection zone and stabilize, thereby avoiding cross-contamination.

We have automated the signal acquisition and basic data analysis. By setting the detection duration for each concentration point and the sample switching interval in the software, the measurement platform will automatically complete the detection of one concentration, rinse the channel, and then trigger the signal acquisition for the next concentration, all without manual intervention.

Software Platform

We have also developed a control platform on a host computer to manage the entire integrated device. This platform can display fluorescence intensity data in the form of dynamic curves, allowing for real-time monitoring by the user.

Our microfluidic testing platform's control software is built on the Streamlit framework, an open-source Python library designed for data scientists and machine learning engineers to rapidly create web applications. Its core features of simplicity, ease of use, integrated functionality, and security and reliability demonstrate significant advantages and broad application prospects.

Within this framework, we have utilized Python libraries to create an interactive system that integrates device control, experimental workflow management, data analysis, and visualization.

The software uses Streamlit's session_state mechanism as its core for global state management, which uniformly stores key information such as pump parameters, experimental process details, system logs, experimental data, and emergency status flags. This ensures page interactivity and state synchronization, providing foundational support for real-time operations and dynamic updates.

The platform lowers the operational barrier through its intuitive, interactive interface, allowing even non-specialists to quickly master device control and experiment management without needing to delve into the underlying technical details. Leveraging its real-time state management mechanism, the platform achieves dynamic synchronization of pump operating status, experimental workflow progress, and data changes, ensuring the experimental process is both monitorable and traceable. It integrates end-to-end functionalities, including device control, process automation, data analysis, and visualization, thereby avoiding the complexity of switching between multiple systems and forming an integrated "control-monitor-analyze" workflow. Furthermore, the emergency stop and state reset mechanisms provide a solid guarantee for experimental safety, enabling rapid responses to unexpected situations.

The platform is suitable for laboratory-scale microfluidic experimental scenarios and can significantly enhance experimental efficiency and reproducibility, particularly in fields like protein reaction detection and microfluidic mixing reactions. In the future, it can be further extended to integrate and control a wider variety of microfluidic devices, supporting more complex customizations of experimental workflows to meet the needs for miniaturized and automated experiments in diverse fields such as biomedical diagnostics, material synthesis, and environmental monitoring. For research institutions or small to medium-sized laboratories, its low-cost and easy-to-deploy nature positions it to become a practical tool for promoting the widespread application of microfluidic technology, facilitating rapid research and development and the translation of results in related fields.

Hardware-Software Interaction for Sample Flow Control

Our custom-designed software plays a crucial role in the precise control of hardware components related to sample handling. Specifically, for the pumps responsible for sample flow within the system, the software allows users to control the flow rates and sequence of both the protein samples and their targets. Through the interface, researchers can set the desired sample flow rates on the pumps, with an adjustable range tailored to the specific requirements of different experiments.

Furthermore, this interaction determines the duration for which the sample flow is maintained, which can be set for a time range from a few seconds to several minutes. Once these parameters are entered into the software, it sends digital signals to the pump's integrated circuit board. The pump's control unit, equipped with a microcontroller, interprets these signals and adjusts the pump's mechanical components. For instance, in the case of a syringe pump, it adjusts the motor's speed or the displacement mechanism to precisely achieve the set flow rate and maintain it for the specified duration. This seamless interaction ensures that samples are delivered to the subsequent detection or reaction zones within the hardware setup at the correct speed and for the appropriate amount of time.

Data Transmission and Affinity Curve Generation

After the sample flows through the relevant channels and interacts with the binding proteins or other detection elements designed in the hardware, the hardware is capable of quantifying affinity based on the principles of FCCS. Two proteins are labeled with fluorescent markers of different colors. When these two markers, as part of the same complex, move in and out of the detection zone together, they cause fluctuations in fluorescence intensity. Synchronized fluctuations appear in both detection channels, resulting in a cross-correlation amplitude that is significantly higher than the background. The magnitude of this amplitude is directly proportional to the number of complexes (i.e., the bound fraction) and can be used to quantify protein interaction and co-diffusion. Through methods like correlation function analysis, this reflects the diffusion characteristics and dynamic changes of the molecules, thereby quantifying their affinity. The hardware's built-in data acquisition module then collects these fluorescence signals and converts the analog signals representing affinity values into digital data that can be processed by the software.

Subsequently, this digital data is transmitted to the software via a wired or wireless connection. Upon receiving the data, the software employs advanced algorithms for data processing and analysis. Based on the affinity data points received over time, the software utilizes mathematical modeling and graphing functions to generate an affinity curve. This curve provides an intuitive visualization of how the affinity between the sample protein and the targeted protein changes throughout the experiment. It allows researchers to quickly and intuitively understand the dynamics of the interaction and to screen for suitable proteins binders based on the curve's results.

In summary, the integration of hardware and software in our integrated architecture forms a closed-loop process: the software controls the sample flow within the hardware, and the hardware feeds the detected data back to the software for further analysis and visualization.

Outlook

Looking ahead, we plan to further refine the chip architecture, exploring to integrate the functions of both the primary and secondary chips into a single microfluidic chip. By achieving gradient generation, sample mixing and detection within one unified chip, we can significantly reduce fluidic connection losses. Meanwhile, we aim to establish a versatile and scalable microfluidic detection platform that not only supports protein screening and characterization in our project, but also provides technical foundations for broader applications in biomolecular interaction studies, drug discovery, and diagnostic development.

Reference

Kirsten Bacia, Petra Schwille. Practical guidelines for dual-color fluorescence cross-correlation spectroscopy, Nature Protocols, 2007, 2(11), 2842-2856

Yu Tian, Michelle M. Martinez, Dimitri Pappas. Fluorescence Correlation Spectroscopy: A Review of Biochemical and Microfluidic Applications, Appl Spectrosc. 2011 April ; 65(4), 115–124

Bor-Shyh Lin, Yu-Ching Yang, Chong-Yi Ho, Han-Yu Yang, Hsiang-Yu Wang. A PDMS-Based Cylindrical Hybrid Lens for Enhanced Fluorescence Detection in Microfluidic Systems, Sensors 2014, 14, 2967-2980

Zhihai Liu, Chao Liu, Yu Zhang , Yaxun Zhang, Xinghua Yang, Jianzhong Zhang , Jun Yang, Libo Yuan. Fiber-Based Helical Channels Refractive Index Sensor Available for Microfluidic Chip, IEEE PHOTONICS TECHNOLOGY LETTERS, 2017, 29(23): 2087-2090

Ren, M., Sun, J., Guan, J., Liu, C., Gong, C., Wang, Y., ... & Protenix Team. (2025). PXDesign: Fast, Modular, and Accurate De Novo Design of Protein Binders. bioRxiv, 2025-08.

Zhu J, Meng Y, Gao W, et al. AI-driven high-throughput droplet screening of cell-free gene expression. Nat Commun. 2025;16(1):2720. Published 2025 Mar 19. doi:10.1038/s41467-025-58139-0

O'Mahoney C, Watt I, Fiedler S, et al. Microfluidic Diffusional Sizing (MDS) Measurements of Secretory Neutralizing Antibody Affinity Against SARS-CoV-2. Ann Biomed Eng. 2024;52(6):1653-1664. doi:10.1007/s10439-024-03478-0

Agu CV, Cook RL, Martelly W, Gushgari LR, Mohan M, Takulapalli B. Multiplexed proteomic biosensor platform for label-free real-time simultaneous kinetic screening of thousands of protein interactions. Commun Biol. 2025;8(1):468. Published 2025 Mar 21. doi:10.1038/s42003-025-07844-z

Amin R, Knowlton S, Hart A, et al. 3D-printed microfluidic devices. Biofabrication. 2016;8(2):022001. Published 2016 Jun 20. doi:10.1088/1758-5090/8/2/022001

Styles MJ, Pixley JA, Wei T, Basile C, Lu SS, Dickinson BC. High-throughput protein binder discovery by rapid in vivo selection. Preprint. bioRxiv. 2025;2025.01.06.631531. Published 2025 Jan 17. doi:10.1101/2025.01.06.631531

Kartal Ö, Andres F, Lai MP, Nehme R, Cottier K. waveRAPID-A Robust Assay for High-Throughput Kinetic Screens with the Creoptix WAVEsystem. SLAS Discov. 2021;26(8):995-1003. doi:10.1177/24725552211013827

Bouzetos E, Ganar KA, van der Oost J, Deshpande S. Cell-free screening of CRISPR-Cas activity by microfluidics-assisted in vitro compartmentalization. Trends Biotechnol. Published online August 14, 2025. doi:10.1016/j.tibtech.2025.07.007