Hardware-Accelerated HMM for Genomic Data Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current bioinformatics methods for analyzing genomic data are labor-intensive, time-consuming, and prone to errors, especially when dealing with the rapid growth of genomic data from Next Gen Sequencers, leading to increased costs and inefficiencies in processing and accuracy.

Innovation Solution

The implementation of a system that uses an integrated circuit with hardwired digital logic circuits to perform bioinformatics protocols such as mapping, alignment, and variant calling, optimizing these processes for faster and more accurate analysis on genetic sequence data, utilizing a hardware-accelerated platform like FPGA or ASIC.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multithreaded software tools are executed on computer clusters with high availability storage, then genomic data analysis can be performed, but processing time and computational costs increase substantially

Engineering Contradiction:
Improvegenomic data processing throughputVSAvoidprocessing time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent replaces software-based multithreaded processing with a hardware-accelerated processing platform that uses dedicated electronic circuits to perform bioinformatics algorithms. This substitution of mechanical/software systems with hardware systems enables parallel processing of genomic data at much higher speeds, directly resolving the contradiction between processing throughput and time consumption.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The processing platform divides genomic data analysis into multiple independent processing channels that can operate simultaneously. Each channel handles specific tasks (mapping, alignment, variant calling) in parallel, increasing overall throughput without proportionally increasing total processing time, thus resolving the contradiction between productivity and time loss.

Inventive Principle:
Principle #1Segmentation

2Reliability

If software-based bioinformatics pipelines are used, then genomic data analysis is achievable, but accuracy and sensitivity are reduced

Engineering Contradiction:
Improveanalysis accuracyVSAvoidcomputational system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces complex software-based analysis pipelines with a hardware-accelerated platform that implements bioinformatics algorithms in dedicated electronic circuits. This hardware implementation provides deterministic processing with consistent accuracy and sensitivity across different runs, while the modular architecture manages complexity through standardized processing units that can be configured for different analysis tasks.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Speed

If hardware-accelerated processing is implemented, then processing speed improves, but device complexity increases

Engineering Contradiction:
Improveprocessing speedVSAvoidhardware system complexity
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The processing platform uses universal processing units that can be configured to perform multiple bioinformatics functions (mapping, alignment, variant calling) through programmable logic. This multi-functionality allows the hardware to achieve high processing speeds for various genomic analysis tasks without requiring separate dedicated hardware for each function, thus managing complexity while maintaining speed improvements.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3608913A1Bioinformatics systems, apparatuses, and methods executed on an integrated circuit processing platform
Publication Date: 2020.02.12 EDICO GENOME CORP
  • EP3608913A1 patent drawingFigure 1~3
  • EP3608913A1 patent drawingFigure 4~6
  • EP3608913A1 patent drawingFigure 7

AI summary

A method for a reduced computation hidden markov model (HMM) in computational biology applications and a system configured to perform the method are provided. The method comprises performing instructions of a HMM pre-filter engine and instructions of a HMM computation engine. The instructions of the HMM pre-filter engine comprise receiving a haplotype sequence at a HMM pre-filter engine; receiving a read sequence at the HMM pre-filter engine; performing a correlation between the haplotype sequence and the read sequence at the HMM pre-filter engine; and generating a plurality of control parameters based on the correlation. The instructions of the HMM computation engine comprise receiving the plurality of control parameters, the read sequence and the haplotype sequence at a HMM computation engine; selecting a reduced number of cells of a HMM matrix structure based on the plurality of control parameters at the HMM computation engine; and computing a HMM metric from the reduced number of cells at the HMM computation engine.