What is the Function of Maximum Likelihood in Phylogenetics?

📚 What is Maximum Likelihood in Phylogenetics?

Maximum likelihood (ML) is a statistical method used to estimate the evolutionary relationships among organisms based on their genetic data (DNA or protein sequences). It aims to find the phylogenetic tree that is most likely to have produced the observed data, given a specific model of sequence evolution.

📜 A Brief History

The concept of likelihood has been around for a while in statistics, but its application to phylogenetics really took off in the late 20th century. Early phylogenetic methods were often based on parsimony (minimizing the number of evolutionary changes), but ML offered a more statistically rigorous approach. Joseph Felsenstein is a key figure in popularizing ML for phylogenetic inference.

🔑 Key Principles Explained

• 🧬 Sequence Alignment: The process starts with aligning the DNA or protein sequences of the organisms you're studying. This alignment shows which positions in the sequences correspond to each other.
• 🌳 Tree Space: ML searches through a vast space of possible phylogenetic trees. Each tree represents a different hypothesis about the evolutionary relationships.
• 📊 Models of Sequence Evolution: These are mathematical models that describe how DNA or protein sequences are expected to change over time. They account for things like different rates of substitution between nucleotides (A, C, G, T). Common models include Jukes-Cantor, HKY, and GTR.

  For example, the Jukes-Cantor model assumes that all nucleotide substitutions occur at the same rate. More complex models, like GTR (General Time Reversible), allow for different rates for each possible substitution.

• 🧮 Likelihood Calculation: For each tree and each model, the likelihood is calculated. The likelihood is the probability of observing the data (the aligned sequences) given that particular tree and model. Mathematically, this is expressed as: $L = P(Data | Tree, Model)$ Where:
  • $L$ is the likelihood
  • $Data$ represents the observed sequence data
  • $Tree$ is a specific phylogenetic tree topology
  • $Model$ is the model of sequence evolution
• 🔎 Optimization: ML algorithms search for the tree and model parameters (e.g., substitution rates) that maximize the likelihood. This often involves complex computational techniques.

  Optimization algorithms like hill-climbing are often used to find the maximum likelihood. These algorithms start with an initial tree and then iteratively modify it until they find a tree with a higher likelihood.

• 💯 Bootstrap Support: To assess the confidence in the resulting tree, a technique called bootstrapping is often used. This involves resampling the original data and re-running the ML analysis multiple times. The percentage of times a particular branch appears in the resulting trees is the bootstrap support value for that branch.

🌍 Real-world Examples

• 🦠 Tracking Virus Evolution: ML is used to trace the evolution of viruses like HIV and influenza. This helps scientists understand how these viruses are spreading and developing resistance to drugs.
• 🐅 Understanding Species Relationships: ML helps to resolve the evolutionary relationships between different species. For example, it can be used to determine how closely related different species of cats are to each other.
• 🌱 Investigating Plant Evolution: ML is essential for studying the evolution of plants, including the origins of important crop species.

🧪 Practical Application of Maximum Likelihood

Here's how the process practically unfolds:

1. Data Collection: Obtain genetic sequences from the organisms of interest.
2. Sequence Alignment: Align these sequences using software like MUSCLE or MAFFT.
3. Model Selection: Choose an appropriate evolutionary model using tools like ModelTest-NG or jModelTest.
4. Tree Search: Use ML software such as RAxML, PhyML, or IQ-TREE to search for the best tree.
5. Tree Evaluation: Assess the robustness of the resulting tree using bootstrap analysis.

🔑 Key Advantages and Disadvantages

• 👍 Advantages:
  • ✅ Statistically sound
  • ✅ Can incorporate complex models of evolution
  • ✅ Provides a measure of confidence (bootstrap support)
• 👎 Disadvantages:
  • ⏱️ Computationally intensive
  • ⚠️ Sensitive to model misspecification (choosing the wrong model)
  • ⚠️ Can be difficult to interpret the results

🎓 Conclusion

Maximum likelihood is a powerful tool for inferring evolutionary relationships. While it can be computationally demanding and requires careful selection of evolutionary models, it provides a statistically rigorous framework for understanding the history of life on Earth. By understanding the principles of ML, researchers can gain deeper insights into the evolutionary processes that have shaped the diversity of organisms around us.