Graduate Defense: Golam Mortuza

Dissertation Information

Title: Robust Digital Nucleic Acid Memory

Program: Doctor of Philosophy in Computing

Advisor: Dr. Tim Andersen, Computer Science

Committee Members: Dr. William Hughes, Materials Science and Engineering; and Dr. Reza Zadegan, Computer Science and Materials Science and Engineering

Abstract

The rapid growth of data generation from electronic devices has created a critical demand for efficient and sustainable data storage solutions. Traditional storage systems face challenges in terms of reliability, energy consumption, and scalability, necessitating the exploration of alternative technologies. This dissertation focuses on digital nucleic acid memory (dNAM) as a potential alternative storage medium and investigates the challenges and potential solutions in the context of dNAM. DNAM utilizes single molecule localization microscopy (SMLM) to encode and store data within Deoxyribonucleic acid (DNA) structures called dNAM origami. SMLM surpasses the limitations of light’s diffraction, enabling the imaging of biological samples at a molecular scale. The robustness and data density of the dNAM algorithm rely heavily on the accuracy and performance of SMLM. Within dNAM, emitter localization and error correction are crucial steps, and this dissertation primarily focuses on these aspects. To improve emitter localization in dNAM, deep learning (DL) techniques are employed. This dissertation investigates the impact of multi-emitter situations, where multiple emitters are attached during data acquisition. A neural network-based image upsampling algorithm is developed to progressively increase the resolution of the image. The algorithm preserves the emitter centroid position while upscaling it to a higher resolution image, effectively isolating attached emitters. By extracting the emitter centroid positions from multiple resolutions, the dissertation analyzes the impact of attached emitters on localization accuracy. Additionally, the dissertation addresses the development of an advanced error correction algorithm for dNAM. A preliminary algorithm is initially used to successfully store 20 bytes of digital information in DNA. However, to improve performance and accuracy, the algorithm is enhanced by incorporating the intensity information of each data point. The impact of this addition is thoroughly studied. Furthermore, the error correction algorithm is extended to support arbitrary-shaped 3D/2D DNA origami structures, enabling scalability and  versatility. The findings of this research highlight the potential of DNA as a viable storage medium and shed light on the challenges and solutions specific to dNAM. The incorporation of DL techniques for emitter localization demonstrates improved accuracy and efficiency. Moreover, the advanced error correction algorithm enhances the reliability and capacity of dNAM. These outcomes contribute to the overall robustness and efficiency of dNAM as a data storage method