Md. Sadik Yasir Tauki - Accelerating DNA Sequence Alignment using Charge-Domain Compute-in-Memory

Accelerating DNA Sequence Alignment using Charge-Domain CIM with Multi-Level FeFETs
Mapping bioinformatics algorithms onto 4-state ferroelectric memory arrays

Project Overview
This project designed a charge-domain compute-in-memory (CIM) architecture using multi-level cell (MLC) Ferroelectric FETs (FeFETs) to accelerate the Needleman–Wunsch (NW) algorithm for DNA sequence alignment. By leveraging 4-state FeFETs (storing A, C, G, T directly as {0,1,2,3}), the design eliminates costly binary packing/unpacking and enables in-memory analog accumulation for alignment scoring .

project report_938663164.pdf

Key Features
🧬 Native 4-State Encoding: Direct storage of DNA bases in FeFET quat-cells
⚡ Charge-Domain CIM: Analog MAC operations with no static power consumption
📊 Memory Optimization: 3–4× reduction in DP matrix storage using multi-level cells
🔄 Parallel Computation: Diagonal/row-wise DP updates with high throughput

Research Contributions

Proposed a 1FeFET–1C quat-cell array for storing DNA bases and DP scores
Adapted NW alignment to in-memory operations, reducing bit-level access overhead
Demonstrated variation-tolerant analog summation across 128 cells, robust under ±4σ Vth shifts

Technical Achievements

4× smaller memory footprint compared to binary-packed storage
128× parallelism in DP updates via row/diagonal reads
Up to 7× energy efficiency improvement over GPU/FPGA accelerators by minimizing data movement

Applications

Bioinformatics accelerators for DNA/RNA sequence alignment
Population-scale genomics requiring compact and energy-efficient hardware
Edge healthcare computing where low-power genome analysis is critical

Impact and Recognition
This work demonstrates how emerging memory devices like FeFETs can be co-designed with algorithms to enable compact, fast, and energy-efficient genomic computing, paving the way for specialized accelerators in precision medicine and large-scale genome sequencing .