Survey on phylogenetic tree construction using machine learning

Abstract

Advances in computational biology have significantly impacted phylogenetic analysis, which is essential for studying evolutionary relationships among species. Phylogenetic inference has traditionally relied on techniques based on statistical models of sequence evolution. Recent developments in Machine Learning (ML) and Deep Learning (DL) offer promising alternatives and enhancements spanning the entire phylogenetic pipeline. This review presents a structured and detailed overview of both classical and machine learning-based methods for Multiple Sequence Alignment (MSA) and phylogenetic inference. We examined current algorithms, tools, and evaluation metrics, categorizing approaches based on their role within the phylogenetic pipeline. A comprehensive visual summary of the entire pipeline is also included, integrating machine learning-driven techniques at every stage. We outline how classical and ML-based methods contribute to different components of the pipeline, with particular attention to recent approaches that bypass traditional alignment using embeddings or end-to-end learning. This review serves as a foundation for understanding current trends and evaluating how emerging machine learning techniques are reshaping phylogenetic inference.

Keywords: Computational biology; Deep learning; Machine learning; Multiple sequence alignment; Phylogenetic tree construction.