Phylogenetically estimated neutral rates and fitness effects of mutations to influenza proteins

By constructing phylogenetic trees from over 100,000 influenza sequences, this study estimates site-specific neutral mutation rates and fitness effects across the viral proteome, revealing significant variation among mutation types, strong cross-viral correlations with SARS-CoV-2 and HIV, and providing a comprehensive, interactive resource for understanding how mutation and selection shape influenza evolution in nature.

The Gist

Imagine the influenza virus as a massive, chaotic library where books (its genetic code) are constantly being copied and rewritten. Sometimes, these rewrites are just random typos that don't change the story at all (neutral mutations). Other times, the typos are so bad they ruin the plot, or so good they make the story a bestseller (mutations that affect fitness).

For a long time, scientists could only look at a few thousand of these books at a time to understand how the library evolves. This new study is like hiring a fleet of super-fast robots to read and organize over 100,000 of these viral books at once. By building a giant family tree from this massive collection, the researchers could finally see the big picture.

Here is what they discovered, broken down into simple concepts:

1. The "Typo" Machine isn't Random
You might think that when the virus makes a mistake, it's equally likely to swap any letter for any other letter (like swapping an 'A' for a 'C' is just as likely as swapping an 'A' for a 'G'). The study found this isn't true. The virus has a very specific "bias" in how it makes mistakes. Some types of typos happen 100 times more often than others. It's as if the virus's copy machine is jammed in a way that prefers certain errors over others.

2. A Family Resemblance with Other Viruses
When the researchers compared these "typo patterns" to other famous viruses like SARS-CoV-2 and HIV, they found a surprising family resemblance. The basic rules of how these viruses make mistakes are very similar, like cousins who all have the same family nose. However, when you look closer at the specific details (like the context of the letters around the typo), the flu virus and SARS-CoV-2 start to look quite different, like cousins who grew up in very different neighborhoods.

3. The "Fitness" Scorecard
The researchers wanted to know: which of these typos actually matter? To figure this out, they played a game of "Expectation vs. Reality."

• The Expectation: Based on the "typo machine" bias they discovered, they calculated how many times a specific mutation should have happened if it didn't matter at all.
• The Reality: They counted how many times that mutation actually appeared in the family tree.
• The Result: If a mutation happened way less often than expected, it means the virus "rejected" it because it was harmful (bad fitness). If it happened as expected, it was likely harmless.

They created a massive scorecard covering about 33,000 harmful-sounding changes and 8,000 harmless-sounding changes across every protein in the flu virus.

4. Hidden Rules and Interactive Maps
This scorecard revealed some surprises. For instance, even changes that were supposed to be "harmless" (synonymous mutations) sometimes showed up less often than expected, suggesting they actually have hidden rules or functions we didn't know about.

To make this huge amount of data easy to explore, the team built interactive heatmaps (like a colorful, clickable map). You can click on any part of the virus's code to see its "fitness score," helping us understand exactly which parts of the virus are fragile and which parts are flexible.

In a Nutshell
This study didn't just look at a few pages of the flu virus's story; it read the whole library. By comparing the virus's natural "mistakes" against what we expect to happen by chance, they created a detailed map of how mutation and selection shape the flu virus in the real world, while also showing how it fits into the broader family of viruses like SARS-CoV-2 and HIV.

Technical Summary

Technical Summary

Problem Statement
The rapid evolution of the influenza virus is driven by the interplay between neutral mutations and natural selection. While phylogenetic methods are well-established for studying these evolutionary processes, their application has historically been constrained by computational and data limitations, typically restricting analyses to only a few thousand influenza genome sequences at a time. This scale limitation hinders the ability to accurately estimate neutral mutation rates and fitness effects across the entire viral genome in a natural setting.

Methodology
To overcome previous scale limitations, the authors constructed phylogenetic trees utilizing a dataset of over 100,000 influenza sequences. These large-scale trees served as the foundation for two primary analytical approaches:

1. Neutral Rate Estimation: The authors calculated neutral rates of mutations for the virus's genome by analyzing the trees. This allowed for the differentiation of rates across 12 distinct nucleotide mutation types (e.g., A-to-C, A-to-G).
2. Fitness Effect Estimation: By comparing the observed frequency of mutations along tree branches against the expected frequency under a neutral model, the authors estimated the fitness effects of mutations. This analysis covered approximately 33,000 nonsynonymous and 8,000 synonymous mutations spanning all influenza proteins.

Key Results

• Variability in Neutral Rates: The study revealed that neutral mutation rates vary significantly, by up to approximately 100-fold, among the 12 different nucleotide mutation types.
• Cross-Viral Comparisons: The estimated neutral rates showed high correlation across influenza, SARS-CoV-2, and HIV. However, the analysis also identified nuanced, context-dependent patterns that exhibited marked differences between influenza and SARS-CoV-2.
• Comprehensive Fitness Mapping: The resulting compendium maps the relationship between sequence and fitness for the entire influenza proteome. Notably, the data highlights regions where synonymous mutations are under functional constraint and provides fitness estimates for proteins that previously lacked extensive experimental measurements.

Significance and Claims
The paper positions this work as a significant step in placing influenza's mutation rates within a broader cross-viral context. By leveraging a dataset an order of magnitude larger than typical studies, the authors claim to deepen the understanding of how mutation and selection shape influenza evolution in nature at a site-specific level. To facilitate further exploration of these findings, the authors have made the data accessible through interactive heatmaps of the estimated fitness effects. The work provides a foundational resource for understanding the evolutionary dynamics of influenza without relying solely on experimental data, particularly for proteins with limited prior characterization.