Sequence Function Prediction Through Information Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the function of symbolic sequences, such as nucleic acid or protein sequences, suffer from noise in data modeling, leading to poor performance when data sets are small, and lack a modeling-free approach to extract functional information.
Innovation Solution
A computational method using information theory to decompose sequence information into contributions from monomers, pairs, triples, etc., discarding noise by leveraging advanced algorithms and statistical techniques, allowing for model-free prediction of sequence functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard modeling approaches are used to determine sequence function, then a model of the sequence-function relationship can be created, but noise in the data is also modeled leading to worsening performance when data sets are small
Solution Approach 1:
The patent segments the sequence information into hierarchical levels (monomers, pairs, triples, etc.) using information decomposition. This allows the method to extract functional information at different orders of correlation while discarding noise, thereby improving prediction accuracy especially when data sets are small.
Solution Approach 2:
The patent extracts only the statistically supported information from sequences by decomposing it into contributions from different levels of monomer interactions. Noise is discarded by leveraging information theory to separate meaningful correlations from random variations, enabling reliable function prediction with limited data.
2Measurement precision
If complex modeling approaches are used to extract functional information from sequences, then function prediction can be achieved, but computational complexity increases
Solution Approach 1:
The patent divides the complex task of function prediction into manageable segments by computing information contributions at different orders (monomers, pairs, triples). This segmentation allows for systematic computation that is less complex than full modeling approaches while maintaining prediction accuracy.
Solution Approach 2:
The patent changes the parameter of analysis from complete sequence modeling to information-theoretic measures (mutual information, correlation coefficients). This parameter transformation simplifies the computational approach by focusing on statistical dependencies rather than complex mechanistic models.
3Reliability
If information decomposition is used to extract only statistically supported information, then noise is discarded, but the method requires advanced algorithms and statistical techniques
Solution Approach 1:
The patent segments the information extraction process into distinct computational steps: calculating mutual information between positions, decomposing into hierarchical levels, and computing correlation coefficients. This segmentation makes the algorithm more tractable while maintaining the ability to discriminate noise from meaningful signals.
Data Source
AI summary
A method of determining the function of a sequence using information decomposition includes providing a plurality of sequences forming a knowledge base, each of the plurality of sequences having respective functions associated therewith, forming a plurality of position weight matrices having different orders based on the sequences, generating a sequence score for each of the plurality of sequences to form a plurality of sequence scores, correlating the respective functions with the sequence scores to form correlation coefficients, selecting a selected order from the different orders based on correlation coefficients, generating a test sequence score for a test sequence based on the selected order and determining a function of the test sequence based on the test sequence score and the knowledge base sequence scores.