QT-GILD

QT-GILD imputes missing quartets in incomplete gene trees arising from sampling biases and biological processes such as gene birth and loss, using unsupervised deep learning to improve species tree estimation in phylogenomics.

Key Features:

• Quartet imputation: Imputes missing quartet relationships to generate a complete quartet distribution from incomplete gene trees.
• Unsupervised deep learning (NLP-inspired): Employs an unsupervised deep learning approach inspired by natural language processing (NLP) to learn and reconstruct missing quartets.
• Accounts for gene tree estimation error: Incorporates considerations of errors inherent in gene tree estimation to refine quartet reconstructions.
• Improves species tree estimation: Enhances species tree accuracy by providing more complete and corrected quartet inputs for downstream inference.
• Simulated and empirical datasets: Applicable to both simulated and empirical phylogenomic datasets.
• Addresses missing-gene scenarios: Targets incomplete gene sets caused by sampling biases and biological phenomena such as gene birth and loss.

Scientific Applications:

• Species tree estimation: Improves inference of species trees from collections of gene trees with missing taxa.
• Phylogenomic analyses with incomplete data: Enables more accurate downstream phylogenomic analyses when gene trees lack a common set of genes.
• Quartet-based methods enhancement: Supplies complete quartet distributions to improve the performance of quartet-based species tree reconstruction methods.

Methodology:

Uses an unsupervised deep learning approach inspired by NLP to learn and reconstruct missing quartets and impute a complete quartet distribution from incomplete gene trees while accounting for gene tree estimation error.