Discriminating between HuR and TTP binding sites using the k-spectrum kernel method.

Area of Science:

• Computational biology
• Molecular biology
• Bioinformatics

Background:

• RNA binding proteins (RBPs) human antigen R (HuR) and Tristetraprolin (TTP) exhibit competitive binding with opposing effects on messenger RNA (mRNA).
• Understanding cellular mechanisms for discriminating between HuR and TTP binding is crucial.
• Machine learning (ML) methods like support vector machines (SVMs) show potential for identifying discriminative features in RBP motifs.

Purpose of the Study:

• To apply ML, specifically SVMs with k-spectrum kernel, to identify discriminative sequence features for HuR and TTP binding.
• To investigate the subtle sequence differences that dictate cellular discrimination between HuR and TTP.
• To explore domain adaptation and multi-task learning for predicting common binding sites.

Main Methods:

• Utilized the k-spectrum kernel with support vector machines (SVMs).
• Employed feature engineering to highlight sequence preferences (U-rich for HuR, AU-rich for TTP).
• Applied domain adaptation and multi-task learning to predict shared binding sites.

Main Results:

• Successfully verified known binding sites for both HuR and TTP.
• Identified distinct sequence preferences: HuR prefers U-rich, TTP prefers AU-rich sequences.
• Demonstrated that increasing A content favors TTP-only binding.

Conclusions:

• Subtle sequence content features distinguish HuR and TTP binding.
• The A/U balance in untranslated regions (UTRs) critically affects RBP binding and mRNA stability.
• K-spectrum kernel and domain adaptation techniques are valuable for studying RBPs with similar binding preferences.