ClaPIM: Scalable Sequence Classification Using Processing-in-Memory

Deoxyribonucleic acid (DNA) sequence classification is a fundamental task in computational biology with vast implications for applications such as disease prevention and drug design. Therefore, fast high-quality classifiers are significantly important. This article introduces ClaPIM, scalable DNA architecture based on the emerging concept of hybrid in-crossbar near-crossbar memristive processing-in-memory (PIM). We enable efficient by uniting filter search stages within single algorithm. Specifically, we propose custom filtering technique that drastically narrows space approach facilitates approximate string matching through distance function. ClaPIM first PIM benefits from high density crossbar arrays massive parallelism PIM. Compared Kraken2, state-of-the-art software classifier, provides higher quality (up to $20 \times $ improvement F1 score) also demonstrates notation="LaTeX">$1.8 throughput improvement. edit tolerant (EDAM), recently proposed static random-access memory (SRAM)-based accelerator restricted small datasets, observe both notation="LaTeX">$30.4 normalized per area 7% increase precision.