[Paper Review] TreeGrafter: phylogenetic tree-based annotation of proteins with Gene Ontology terms and other annotations

Haiming Tang, ROBERT FINN+1 more|arXiv (Cornell University)|Feb 20, 2018

Genomics and Phylogenetic Studies5 references4 citations

TL;DR

TreeGrafter is a computational tool that enhances protein annotation by grafting query sequences onto pre-annotated phylogenetic trees and propagating functional annotations—such as Gene Ontology terms and PANTHER classifications—from ancestral nodes. It outperforms subfamily HMM scoring in subfamily assignment and delivers highly specific GO annotations by leveraging evolutionary relationships in reference trees.

ABSTRACT

Summary: TreeGrafter is a new software tool for annotating protein sequences using annotated phylogenetic trees. Cur-rently, the tool provides annotations to Gene Ontology terms, and PANTHER protein class, family and subfamily. The ap-proach is generalizable to any annotations that have been made to internal nodes of a reference phylogenetic tree. Tree-Grafter takes each input query protein sequence, finds the best matching homologous family in a library of pre-calculated, pre-annotated gene trees, and then grafts it to the best location in the tree. It then annotates the sequence by propagating anno-tations from its ancestral nodes in the reference tree. We show that TreeGrafter outperforms subfamily HMM scoring for cor-rectly assigning subfamily membership, and that it produces highly specific annotations of GO terms based on annotated reference phylogenetic trees. This method will be further inte-grated into InterProScan, enabling an even broader user com-munity. Availability: TreeGrafter is freely available on the web at https://github.com/haimingt/TreeGrafting.

Motivation & Objective

• To develop a method for transferring functional annotations from annotated phylogenetic trees to novel protein sequences using evolutionary relationships.
• To improve the accuracy of subfamily membership assignment in protein families compared to existing HMM-based approaches.
• To enable specific and biologically meaningful Gene Ontology term annotations by leveraging ancestral node annotations in reference trees.
• To integrate the method into InterProScan for broader accessibility across the bioinformatics community.
• To generalize the approach to any annotations assigned to internal nodes of a reference phylogenetic tree.

Proposed method

• The tool identifies the best-matching homologous family in a pre-computed library of annotated gene trees for each query protein sequence.
• It performs sequence alignment to determine the optimal grafting position within the reference tree based on evolutionary distance.
• Annotations are propagated from the ancestral nodes of the reference tree to the query sequence along the path from the graft point to the root.
• The method relies on pre-annotated reference phylogenetic trees where internal nodes are already labeled with functional terms such as GO terms and PANTHER classifications.
• The approach uses sequence similarity and phylogenetic placement to infer functional annotations without requiring de novo HMM modeling.
• The tool is designed to be extensible to any annotation type assigned to internal nodes of a reference tree, not limited to GO or PANTHER.

Experimental results

Research questions

• RQ1 Can phylogenetic tree-based annotation improve the accuracy of subfamily membership prediction compared to subfamily HMM scoring?
• RQ2 To what extent does tree-based annotation transfer Gene Ontology terms with higher specificity than sequence-based methods?
• RQ3 How effectively can ancestral annotations from reference trees be propagated to novel query sequences through phylogenetic grafting?
• RQ4 Can the method be generalized to other types of annotations beyond GO and PANTHER classifications?
• RQ5 What is the performance of TreeGrafter in comparison to existing tools like InterProScan in functional annotation transfer?

Key findings

• TreeGrafter outperforms subfamily HMM scoring in correctly assigning subfamily membership, demonstrating improved precision in functional classification.
• The method produces highly specific Gene Ontology annotations by leveraging the evolutionary context encoded in annotated reference trees.
• TreeGrafter is designed for integration into InterProScan, enabling broader adoption across the bioinformatics community.
• The approach is generalizable to any annotations assigned to internal nodes of a reference phylogenetic tree, not limited to GO or PANTHER terms.
• The tool achieves accurate functional annotation by propagating annotations from ancestral nodes, ensuring biologically coherent inferences based on evolutionary relationships.