Statistical Design of a 3D Microarray with Position-Encoded Microspheres
Arye Nehorai


Microarrays are used to analyze concentrations of multiple targets (proteins, antibodies, mRNA, etc.) in parallel. Applications include medical screening, drug discovery, gene sequencing, self-administered medical testing (pregnancy tests, glucose meters, and cholesterol meters), environmental monitoring, and homeland security. A three-dimensional (3D) microarray detects bio-targets employing quantum-dot (QD) barcoded polystyrene microspheres containing molecular probes. It uses optical reporters (e.g., fluorescent dyes, QDs, nanocrystals) to identify the targets. In existing 3D microarrays, the microspheres are positioned randomly within a substrate, resulting in inefficiently packed microspheres and time-consuming image analysis.

In this talk, I will present an overview of our research on designing and imaging of 3D microarrays. I will first discuss our statistical approach to optimally designing the layout of microspheres with controllable positions in a 3D microarray. The controllable positions enable optimizing the packing of the microspheres and encoding based on position. We employ the posterior Cram?r-Rao bound to ensure the desired statistical imaging performance. The new (position-encoded) microarray has high sensitivity, efficient packing density, and guaranteed statistical imaging performance.

I will compare the performance of the new microarrays with the existing ones (that use random positions), for automated image analysis. To perform such analysis for the new microarray, we conduct a grid-based segmentation of its microspheres' images. For the existing microarray, we achieve the segmentation by considering the local convex nature of the microspheres and using an edge detection scheme; we identify the targets using a maximum-likelihood (ML) deblurring method and an M-ary testing analysis. For both types of arrays, we quantify the target levels using an ML estimation considering a sparse target-intensity profile.

Finally, I will present collaborative results on microfluidic fabrication of the position-encoded microarrays. Verifying the performance improvements, we envision employing the proposed device for example to malaria diagnostics in the third world, where existing techniques are too bulky and expensive for fast on-site detection.

Note: This work has been done in collaboration with Pinaki Sarder, PhD student, ESE department, WUSTL.