High-throughput combinatorial screening of antiplatelet drugs for personalized medicine

Abstract

Cardiovascular disease (CVD) remains the leading cause of death worldwide. Platelet activation plays a critical role in arterial thrombotic events such as myocardial infarction. Although antiplatelet drugs are standard therapies, they are associated with risks including bleeding, gastrointestinal adverse effects, and drug resistance. Furthermore, substantial inter-individual variability in patient responses underscores the need for personalized antiplatelet regimens. These factors emphasize the importance of screening for optimal antiplatelet drugs and drug combinations tailored to individual patients. However, traditional platelet detection assays are reagent-hungry and low-throughput, making them unsuitable for high-throughput screening of antiplatelet agents. Here, we present the C-chip, a high-throughput platform for on-chip parallel screening of antiplatelet drug combinations. The C-chip miniaturizes individual screening reactions into picoliter-volume, color-coded droplets, enabling the generation of thousands of screening data points in a single experiment. We demonstrate that the C-chip can effectively identify the optimal combinations of three clinically relevant antiplatelet drugs: Aspirin, Tirofiban, and Ticagrelor. We further applied this platform to identify optimal drug combinations for five healthy volunteers, revealing marked inter-individual variability in antiplatelet drug responses.

Fig. 1. Overview of the combinatorial screening chip (C-chip).

a Fluorescent barcode drug droplets and platelet droplets with varying diameters were generated using microfluidics. The drugs were pre-mixed with different fluorescent markers. b Platelet droplets were loaded into the microwell array device and occupy the larger microwells, followed by the loading of drug droplets that occupy the smaller microwells. Each small microwell randomly captured one drug droplet. Through corona treatment, the droplets in each microwell unit were merged, enabling combinatorial drug treatment for platelets and fluorescent labeling. c Using a fluorescence microscope, the aggregation and dispersion of platelets were continuously observed and recorded, with real-time images captured for each area of the chip. Subsequently, decoding software was used to re-identify the platelets and calculate the anti-aggregation efficiency of platelets for each individual drug or drug combination

Fig. 2. Optimization of platelet microdroplet generation.

a Micrographs of the generated platelet droplets and drug droplets. The scale bar for all images is 50 μm. b Diameter distribution of platelet droplets (n = 500) and drug droplets (n = 499). c Platelet aggregation of washed control (without corona treatment, n = 6) and corona-treated platelets (n = 6) measured by aggregometry. d Platelet aggregation of washed control (n = 6) and fluorophore-treated platelets (n = 6) measured by aggregometry. e Platelet droplets exposed to various thrombin concentrations (0, 0.01, 0.05, 0.1, 0.2, 0.5, 1, and 5 U/mL). f Variation in the number of platelets within droplets after exposure to a constant thrombin concentration. The data are expressed as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. 0 U/mL group

Fig. 3. The performance of the C-chip and optimization of the AI-based aggregation evaluation model.

a Micrographs showing the process of blank droplet loading, drug droplet loading and droplet merging. Scale bar for all images is 100 μm. b AI-based platelet aggregation evaluation model trained via transfer learning from a general image classification network. The model predicted the probability distribution of platelet states within microwells after decoding the fluorescent barcodes. It classified four distinct states: (1) droplet with aggregated platelets (clustered); (2) droplet with dispersed platelets (dispersed); (3) droplet present but no platelets (drop wo_cell); and (4) microwell without any droplet (wo_drop). c Training and validation accuracy of the AI-based platelet aggregation model. d Training and validation loss of the AI-based platelet aggregation model

Fig. 4. IC50 evaluation of different antiplatelet drugs on the C-chip.

a The concentrations of Aspirin tested were 0, 0.2, 0.4, 2, 2.2, 4, 20, 20.2, 22, and 40 μM. b Nonlinear fitting curve for determining the IC50 value (2.149 μM) of Aspirin. c The concentrations of Tirofiban tested were 0, 0.1, 0.2, 1, 1.1, 2, 10, 10.1, 11, and 20 μM. d Nonlinear fitting curve for determining the IC50 value (0.0997 μM) of Tirofiban. e The concentrations of Ticagrelor tested were 0, 0.1, 0.2, 1, 1.1, 2, 10, 10.1, 11, and 20 μM. f Nonlinear fitting curve for determining the IC50 value (0.9242 μM) of Ticagrelor. The data were expressed as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. 0 μM group

Fig. 5. Combinatorial screening of antiplatelet drugs on the C-chip.

a Heatmap of drug combination tests for volunteer 1. b Anti-aggregation efficiency of single-drug combinations for platelets in the combinatorial screening tests for Volunteer 1. For Aspirin (Asp), the concentrations were 2 μM, 4 μM (2 μM + 2 μM), 20 μM, 22 μM (2 μM + 20 μM), and 40 μM (20 μM + 20 μM). For Tirofiban (Tif), the concentrations were 0.2 μM, 0.4 μM (0.2 μM + 0.2 μM), 2 μM, 2.2 μM (0.2 μM + 2 μM), and 4 μM (2 μM + 2 μM). For Ticagrelor, the concentrations were 1 μM, 2 μM (1 μM + 1 μM), 10 μM, 11 μM (10 μM + 1 μM), and 20 μM (10 μM + 10 μM). c Anti-aggregation efficiency of two-drug combinations for platelets in the combinatorial screening tests for Volunteer 1. Based on the concentration of drug combinations, the two-drug combinations were divided into three groups: low concentration group (low group) including 2 μM Asp + 0.2 μM Tif, 2 μM Asp + 1 μM Tig, 0.2 μM Tif + 1 μM Tig; high concentration group (high group) including 20 μM Asp + 2 μM Tif, 20 μM Asp + 10 μM Tig, 2 μM Tif + 10 μM Tig; combinations of low and high concentrations (medium group) including 2 μM Asp + 10 μM Tig; 2 μM Tif + 1 μM Tig; 2 μM Asp + 2 μM Tif; 20 μM Asp + 1 μM Tig; 20 μM Asp + 0.2 μM Tif; 0.2 μM Tif +10 μM Tig. d Total heatmap of the anti-aggregation efficiency of two-drug combinations for the five volunteers. e The therapeutic effects of different drug combinations across various samples were compared, with samples having an anti-aggregation degree greater than 50% scored as 1 (indicated in orange) and those less than 50% scored as 0 (indicated in gray). f The top three drug combinations with the highest anti-aggregation efficiency in the medium group for the five volunteers