HERP: Hardware for Energy Efficient and Realtime DB Search and Cluster Expansion in Proteomics

2025-11-05

AI Summary

To address the high computational overhead, latency, and low energy efficiency of proteomics database search and clustering under resource-constrained conditions, this paper proposes a lightweight incremental clustering and highly parallelized search architecture. Methodologically, it leverages mass spectrometry data characteristics to design a pre-clustering-guided incremental clustering strategy, integrated with bucket-level parallelism and dynamic query scheduling. The architecture is implemented on a 7 nm 3T 2M TJ SOT-CAM in-memory computing hardware platform to enable compute–memory co-acceleration. Experiments on the human reference proteome (131 GB) show that initializing 2 million spectra consumes only 1.19 mJ, while 1,000 queries consume just 1.1 μJ. Compared to baseline methods, clustering error increases by only 0.3%, search throughput improves 20×, hardware parallelism reaches 100×, and result overlap with mainstream tools is 96%, demonstrating superior trade-offs among accuracy, speed, and energy efficiency.

Technology Category

• Search and Optimization: Distributed Search
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Machine Learning: Hardware-aware ML

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Database (DB) search and clustering are fundamental in proteomics but conventional full clustering and search approaches demand high resources and incur long latency. We propose a lightweight incremental clustering and highly parallelizable DB search platform tailored for resource-constrained environments, delivering low energy and latency without compromising performance. By leveraging mass-spectrometry insights, we employ bucket-wise parallelization and query scheduling to reduce latency. A one-time hardware initialization with pre-clustered proteomics data enables continuous DB search and local re-clustering, offering a more practical and efficient alternative to clustering from scratch. Heuristics from pre-clustered data guide incremental clustering, accelerating the process by 20x with only a 0.3% increase in clustering error. DB search results overlap by 96% with state-of-the-art tools, validating search quality. The hardware leverages a 3T 2M T J SOT-CAM at the 7nm node with a compute-in-memory design. For the human genome draft dataset (131GB), setup requires 1.19mJ for 2M spectra, while a 1000 query search consumes 1.1{mu}J. Bucket-wise parallelization further achieves 100x speedup.

Problem

• Enabling efficient proteomics database search in resource-limited environments
• Reducing energy consumption and latency for real-time clustering
• Achieving high-speed incremental clustering with minimal accuracy loss

Innovation

• Lightweight incremental clustering for proteomics data
• Bucket-wise parallelization to reduce search latency
• Compute-in-memory hardware design for energy efficiency