[Paper Review] Reconstructing protein and gene phylogenies by extending the framework of reconciliation

Esaie Kuitche, Lafond, Manuel+1 more|arXiv (Cornell University)|Oct 30, 2016

Genomics and Phylogenetic Studies15 references3 citations

TL;DR

This paper introduces a novel reconciliation framework that distinguishes gene trees from protein trees, modeling protein evolution through creation and loss events alongside standard gene-species events. It proposes two new problems—MinDRGT and MinDRPGT—to reconstruct accurate gene and protein phylogenies by minimizing a double reconciliation cost with species and protein trees, demonstrating significant cost reductions in Ensembl gene tree corrections using a heuristic algorithm.

ABSTRACT

The architecture of eukaryotic coding genes allows the production of several different protein isoforms by genes. Current gene phylogeny reconstruction methods make use of a single protein product per gene, ignoring information on alternative protein isoforms. These methods often lead to inaccurate gene tree reconstructions that require to be corrected before being used in phylogenetic tree reconciliation analyses or gene products phylogeny reconstructions. Here, we propose a new approach for the reconstruction of accurate gene trees and protein trees accounting for the production of alternative protein isoforms by the genes of a gene family. We extend the concept of reconciliation to protein trees, and we define a new reconciliation problem called MinDRGT that consists in finding a gene tree that minimizes a double reconciliation cost with a given protein tree and a given species tree. We define a second problem called MinDRPGT that consists in finding a protein tree and a gene tree minimizing a double reconciliation cost, given a species tree and a set of protein subtrees. We provide algorithmic exact and heuristic solutions for some versions of the problems, and we present the results of an application to the correction of gene trees from the Ensembl database. An implementation of the heuristic method is available at https://github.com/UdeS-CoBIUS/Protein2GeneTree.

Motivation & Objective

• To address inaccuracies in current gene tree reconstructions that ignore alternative protein isoforms produced by eukaryotic genes.
• To formalize a two-layer reconciliation model where protein trees evolve within gene trees, extending classical gene-species tree reconciliation.
• To develop new optimization problems—MinDRGT and MinDRPGT—that jointly minimize reconciliation costs between protein, gene, and species trees.
• To provide exact and heuristic algorithms for solving these problems, particularly for correcting existing gene trees from databases like Ensembl.
• To evaluate the method on real data, showing that corrected gene trees achieve significantly lower reconciliation costs.

Proposed method

• Introduces protein-gene tree reconciliation by defining two new evolutionary events: protein creation and protein loss, in addition to speciation, duplication, and loss.
• Proposes the MinDRGT problem: find a gene tree minimizing the sum of protein-gene and gene-species reconciliation costs, given a protein tree and species tree.
• Proposes the MinDRPGT problem: jointly reconstruct a protein tree and gene tree minimizing the same double reconciliation cost, given a species tree and a set of protein subtrees.
• Develops a heuristic algorithm for MinDRGT that modifies an input protein tree using the species tree as a guide to produce an optimal gene tree.
• Reduces MinDRPGT to MinDRGT by using a soft-clustering approach to identify creation-free protein subtrees, then reconstructs the full protein tree via a polynomial-time algorithm.
• Applies the heuristic to correct 10,861 Ensembl gene trees, with results showing substantial cost reductions and increased accuracy.

Experimental results

Research questions

• RQ1 Can gene trees be reconstructed more accurately by explicitly modeling protein isoform production through protein creation and loss events?
• RQ2 Is the optimal gene tree necessarily identical to the input protein tree when each gene produces a single protein, or can it differ to minimize reconciliation cost?
• RQ3 Can the joint reconstruction of protein and gene trees (MinDRPGT) be solved efficiently when given only the set of maximum creation-free protein subtrees?
• RQ4 What is the computational complexity of MinDRGT and MinDRPGT, and can they be approximated or solved exactly under specific constraints?
• RQ5 How effective is the proposed method in reducing reconciliation cost on real biological datasets like Ensembl?

Key findings

• Among 10,861 Ensembl gene trees, 96.73% of large trees (100–199 leaves) were modified by the heuristic algorithm, indicating widespread suboptimality in initial reconstructions.
• Modified trees showed a 15.84 ms average reduction in running time and a 4.35% average reduction in double reconciliation cost, with 307.29 average cost reduction per tree.
• The heuristic method reduced the double reconciliation cost by 18.67% on average for small trees (1–9 leaves), 6.01% for medium trees (10–99 leaves), and 4.35% for large trees (100–199 leaves).
• Even when each gene produces only one protein, the optimal gene tree may differ from the input protein tree, showing that gene and protein phylogenies are not equivalent.
• The MinDRPGT problem can be reduced to MinDRGT in polynomial time when all maximum creation-free protein subtrees are known, enabling exact reconstruction of the full protein tree.
• The algorithm successfully reconstructed the full protein tree from its creation-free subtrees with $ O(n^3) $ time complexity, validating the theoretical reduction.