The study of how different cell types evolved across species is key to understanding life’s diversity and unlocking the secrets of biology. One powerful tool for this is single-cell RNA sequencing (scRNA-seq), which allows scientists to examine gene expression in individual cells. By applying this technology across species, researchers can compare and contrast cellular function and evolution in ways never before possible. However, cross-species scRNA-seq studies come with significant challenges, such as genetic differences between species, experimental variability, and the natural biological diversity between individuals.

To tackle these issues, researchers at KAUST compared nine data-integration methods on a massive dataset of 4.7 million cells from 20 species, spanning eight major animal groups (phyla). Their goal? To determine the best tools for combining data across species while minimizing errors and preserving biological meaning.

Why Is Cross-Species Data Integration Challenging?

When analyzing data from different species, several obstacles arise:

1. Genetic Differences: Each species has unique genetic features that can make direct comparisons tricky.
2. Experimental Variations: Differences in how samples are prepared or analyzed can introduce “batch effects,” or artificial discrepancies in the data.
3. Biological Variation: Natural differences between individual organisms can make it hard to distinguish meaningful patterns from noise.

To address these issues, integration methods are used. These tools aim to align data from different species in a way that allows for meaningful comparisons while minimizing distortions.

Schematic diagram of the benchmarking workflow

Here, nine data integration methods were tested using 36 cross-species integration tasks. Integration results were evaluated using 13 metrics that assess batch effect removal (species mixing), nested batch effect removal and conservation of biological variance. The influence of imbalanced dataset and data sequencing depth on the methods were also assessed.

The Study’s Key Findings

The study evaluated nine integration methods to see how well they balanced two goals:

• Removing Batch Effects: Eliminating artificial discrepancies caused by experimental differences.
• Preserving Biological Variance: Retaining meaningful differences that reflect evolutionary and functional relationships.

Here’s what they found:

• Gene Sequence-Based Methods: Tools that incorporate gene sequence information excel at preserving biological variance. They’re ideal for studies exploring evolutionary relationships.
• Generative Models: These methods are great at removing batch effects, ensuring cleaner datasets for downstream analysis.

Standout Tools and Their Strengths

Among the methods tested, three stood out for their specific strengths:

1. SATURN: This tool performed well across a wide range of taxonomic levels, from comparisons between closely related genera to vastly different phyla. Its versatility makes it a great general-purpose choice for integrating diverse datasets.
2. SAMap: Best for large-scale projects involving very distantly related species, SAMap excels in atlas-level integration, offering insights into species beyond the cross-family level.
3. scGen: For studies within more closely related groups (such as within a class), scGen provides strong integration capabilities.

Practical Recommendations

The study provides valuable guidelines for researchers choosing integration methods based on their goals:

• For broad evolutionary studies, tools like SATURN and SAMap are excellent choices.
• For studies focused on closely related species, scGen may be more appropriate.
• When balancing batch effect removal and biological variance preservation, it’s crucial to select the method that aligns best with the specific dataset and research question.

Why is This Important?

Understanding how cell types evolve and function across species has profound implications. It can lead to insights into disease mechanisms, developmental biology, and even the discovery of new therapeutic targets. By fine-tuning the methods used to analyze cross-species scRNA-seq data, scientists can push the boundaries of what we know about life’s diversity and interconnectedness.

This study not only highlights the potential of cross-species single-cell RNA-seq but also offers a roadmap for researchers navigating this exciting and complex field. With the right tools, we can continue to uncover the shared and unique aspects of life across the animal kingdom.