Machine learning method using position-specific mutation based classification outperforms one hot coding for disease...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Haemophilia A is an X-linked genetic disorder caused by deficiencies in clotting factor VIII.
• Disease severity is linked to specific mutations, but predicting this accurately is challenging.
• Existing machine learning methods often suffer from high time complexity and reduced accuracy.

Purpose of the Study:

• To develop and evaluate a novel Position-Specific Mutation (PSM) encoding technique for predicting Haemophilia A mutational severity.
• To compare the performance of PSM against traditional One-Hot Encoding (OHE) in machine learning models for Haemophilia A.

Main Methods:

• A dataset of 7784 Haemophilia A mutations was processed using both PSM and OHE encoding techniques.
• Machine learning algorithms were trained and tested on the encoded datasets for mutation severity classification.
• Performance was evaluated based on prediction accuracy and computational time efficiency.

Main Results:

• The Position-Specific Mutation (PSM) method demonstrated superior performance compared to One-Hot Encoding (OHE).
• PSM achieved significant improvements in training and prediction time, ranging from 91-98% and 80-99% respectively.
• Enhanced accuracy in severity prediction was observed across various machine learning algorithms when using PSM.

Conclusions:

• PSM is a more efficient and accurate encoding technique for predicting Haemophilia A mutation severity than OHE.
• This advancement offers potential for improved diagnostic tools and personalized treatment strategies for Haemophilia A patients.
• The findings highlight the importance of advanced encoding methods in machine learning applications for genetic disorders.