DNA-CIM: DNA Sequence Analysis Using RRAM-Based Compute In-Memory Accelerator

Le résumé fourni par la source

DNA sequence matching and mutation analysis are critical computational methods used in modern genomics, with applications ranging from identifying genetic disorders to under-standing evolutionary relationships. Moreover, understanding the functional effects of genetic modifications through mutation analysis can help in understanding disease mechanisms, treatment resistance, and evolutionary processes. The computational complexity of DNA sequence analysis, which involves huge amounts of data, causes memory wall problem in conventional computing methods. To tackle these issues, in this paper, we introduce a novel Compute-in-Memory (CIM) architecture utilizing Resistive Random Access Memory (RRAM) to perform pairwise DNA sequence matching and mutation analysis tasks efficiently. Our approach converts DNA sequences into binary digits, enabling in-memory bitwise XOR and AND operations using RRAM macro and Multi Logic Sense Circuits (MLSC) for efficient sequence alignment. Additionally, near-memory comparison logic is utilized to classify various mutations within the DNA sequences accurately. The proposed architecture, simulated using TSMC 65nm technology, demonstrates significant energy efficiency of 2.042 TCUPS/W and achieves a throughput of 38.4 GCUPS. By leveraging the inherent advantages of CIM, our architecture offers high parallelism, reduced latency, and enhanced scalability while also addressing challenges such as architecture variability.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.