Metagenomic Sequence Classification via Compositional Clustering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional taxonomic classification techniques for metagenomic sequences are either not robust enough for accurate classification of short sequences, leading to non-specific assignments, or require substantial computational resources due to extensive similarity-based comparisons, limiting their efficiency and accuracy.

Innovation Solution

The method involves creating reference clusters based on compositional characteristics, selecting a target cluster for a query sequence through composition-based analysis, and performing similarity-based analysis only within that cluster, reducing computational time and resources while maintaining accuracy and specificity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional similarity-based classification techniques are used, then classification accuracy is improved, but computational time and resources increase substantially

Engineering Contradiction:
Improveclassification accuracyVSAvoidcomputational time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the reference sequence database into multiple clusters based on compositional characteristics (e.g., k-mer frequencies, GC content). This segmentation allows the query sequence to be compared only against relevant clusters rather than the entire database, reducing computational time while maintaining classification accuracy through targeted similarity-based analysis within each cluster.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary composition-based analysis to identify candidate clusters before conducting similarity-based analysis. This preliminary action filters the search space by pre-grouping reference sequences into compositional clusters, allowing the subsequent similarity-based classification to focus only on relevant subsets and thereby reducing overall computational time and resources.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If composition-based analysis is used, then computational efficiency is improved, but classification accuracy decreases leading to non-specific assignments

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidclassification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent merges composition-based analysis and similarity-based analysis into a hybrid classification approach. Composition-based clustering is used to efficiently organize the reference database and identify candidate clusters, while similarity-based analysis is applied within those clusters to ensure accurate and specific taxonomic assignments. This combination maintains computational efficiency while improving classification accuracy.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent applies different analysis methods to different parts of the classification process: composition-based analysis is used for database organization and candidate identification (where speed is prioritized), while similarity-based analysis is applied locally within candidate clusters (where accuracy is prioritized). This local quality approach ensures each method is used where it provides the most value.

Inventive Principle:
Principle #3Local quality

3Manufacturing precision

If extensive similarity-based comparisons are performed, then classification specificity is improved, but device complexity and resource requirements increase

Engineering Contradiction:
Improveclassification specificityVSAvoidsystem complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the reference database into compositional clusters to reduce the scope of similarity-based comparisons. By organizing sequences into clusters based on compositional characteristics, the system reduces device complexity and resource requirements while maintaining classification specificity through targeted similarity analysis within each cluster rather than across the entire database.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2390810B1Taxonomic classification of metagenomic sequences
Publication Date: 2019.10.16 TATA CONSULTANCY SERVICES LTD
  • EP2390810B1 patent drawingFigure 1
  • EP2390810B1 patent drawingFigure 2
  • EP2390810B1 patent drawingFigure 3

AI summary

Method(s) for identifying a taxon corresponding to a query sequence are described herein. The method includes selecting a target cluster, from amongst a plurality of reference clusters (165), corresponding to the query sequence. The target cluster may be selected based on a composition based analysis. A similarity based analysis of the query sequence is performed with respect to the target cluster. From the target cluster, the taxon corresponding to the query sequence is identified based on the similarity based analysis.