CSI-SSU: Phylogenetic contamination screening of genomic datasets, demonstrated on the Protist 10,000 Genomes (P10K) database

The authors present CSI-SSU, a scalable command-line tool that utilizes SSU phylogenetic placement and bacterial BUSCO searches to screen genomic assemblies for contamination and validate taxonomic identities, as demonstrated by its application to 2,960 assemblies from the Protist 10,000 Genomes (P10K) initiative.

The Gist

Imagine you are trying to build a massive, perfect library of every living thing on Earth, specifically focusing on the microscopic "protists" (tiny single-celled organisms) that are often overlooked. This is the goal of a huge project called P10K (Protist 10,000 Genomes). They want to sequence the DNA of 10,000 different species to understand how life evolved.

But there's a big problem: The library is messy.

Because these tiny organisms live in mud, water, and soil, they are often surrounded by other bugs, fungi, and bacteria. When scientists try to sequence their DNA, they accidentally grab the DNA of the neighbors, too. It's like trying to take a photo of a single flower in a garden, but your camera accidentally snaps pictures of the grass, the bugs, and the fence posts right next to it. If you don't clean this up, your "flower photo" is actually a confusing collage of everything in the garden.

Enter the Detective: CSI-SSU

The authors of this paper created a new digital tool called CSI-SSU (Contaminant Sequence Investigation). Think of this tool as a super-smart, automated librarian with a magnifying glass and a DNA fingerprinting kit.

Here is how it works, using simple analogies:

1. The "Name Tag" Check (SSU Sequences)

Every organism has a specific "name tag" in its DNA called the SSU (a small piece of ribosomal RNA). It's like a barcode on a product in a grocery store.

• The Problem: When the P10K project gets a DNA sample, it might have the barcode for a "Slime Mold" mixed with barcodes for "Mushrooms" and "Beetles."
• The CSI-SSU Solution: The tool scans the DNA pile, finds all the barcodes, and checks them against a giant, curated list of known organisms (the PR2 database). It instantly says, "Hey, this barcode belongs to a Slime Mold, but this other one belongs to a Beetle. The Beetle doesn't belong here!"

2. The "Fake ID" Detector (Chimeras)

Sometimes, DNA gets scrambled during the sequencing process, creating a "Frankenstein" sequence that is half-organism A and half-organism B.

• The CSI-SSU Solution: The tool acts like a bouncer at a club checking IDs. It looks for these "chimeric" (mixed-up) sequences and flags them as fake or broken, so scientists know to throw them out.

3. The "Guest List" Verification (Phylogenetic Placement)

This is the most clever part. Instead of just matching a barcode to a name, the tool places the DNA on a family tree.

• The Analogy: Imagine you find a stranger at a family reunion. You don't just ask, "Who are you?" You look at the family tree. If the stranger is standing next to the "Cousins from Ohio," but the DNA says they are from "Cousins from Texas," the tool knows something is wrong.
• The Result: CSI-SSU places the DNA on the evolutionary tree. If a sample supposed to be a "Slime Mold" ends up sitting on the branch of "Fungi," the tool screams, "Contamination!"

4. The "Bacterial Sniffer" (BUSCO)

The tool also checks for bacteria. It looks for specific bacterial genes that shouldn't be there. If it finds too many, it knows the sample is heavily contaminated with bacteria, even if the main organism is a protist.

What Did They Find?

The authors tested this tool on 2,960 samples from the P10K database. It was like doing a massive audit of the library.

• The Messy Truth: They found that contamination is everywhere. Many samples that were thought to be pure "Slime Molds" were actually mixed with fungi, plants, or other protists.
• The Misunderstandings: Some organisms were mislabeled entirely. For example, a sample thought to be one type of amoeba was actually a completely different type. The tool corrected these mistakes by looking at the family tree.
• The "Good" vs. "Bad" Samples: The tool helped separate the "High-Quality" samples (clean, pure DNA) from the "Low-Quality" ones (messy, contaminated DNA). This tells scientists which samples they can trust for future research and which ones need to be re-sequenced or cleaned up.

Why Does This Matter?

Imagine trying to write a history book about the evolution of life, but your sources are full of lies and mixed-up facts. You would get the wrong story.

By using CSI-SSU, scientists can now:

1. Trust the Data: They know exactly which samples are clean.
2. Fix Mistakes: They can correct the names of organisms that were labeled wrong.
3. Save Time: Instead of manually checking thousands of DNA files (which would take years), this tool does it in minutes.

The Bottom Line

The P10K project is a fantastic effort to map the microscopic world, but it was a bit of a "wild west" with lots of messy data. CSI-SSU is the sheriff that came in, organized the files, kicked out the impostors, and made sure the library is ready for serious study. It ensures that when we learn about the history of life on Earth, we are learning from the truth, not from a mix-up of DNA from the garden next door.

Technical Summary

1. Problem Statement

The rapid expansion of genomic resources for microbial eukaryotes (protists) is hindered by three critical challenges:

• Taxonomic Imprecision: Reliance on BLAST-based similarity searches for identification often yields less accurate classifications compared to phylogeny-based methods, leading to misidentified target lineages.
• Widespread Contamination: Protists often inhabit complex microbial communities, host endosymbionts, or prey on other organisms. Consequently, genomic assemblies frequently contain non-target eukaryotic sequences (cross-group contamination) and bacterial/archaeal DNA. Current decontamination pipelines often fail to remove non-target eukaryotes effectively.
• Data Reproducibility: Large-scale databases like the Protist 10,000 Genomes (P10K) often lack detailed metadata (e.g., sequencing platforms, voucher images), making it difficult to assess data quality or reproduce results.

The authors argue that robust, scalable, and phylogenetically informed tools are essential to validate taxonomic identities and screen for contamination before these datasets can be used for evolutionary or functional genomics.

2. Methodology

The authors developed CSI-SSU (Contaminant Sequence Investigation using Small Subunit ribosomal RNA), a command-line tool designed to automate the screening, retrieval, and classification of SSU (18S) sequences from genomic assemblies.

Core Workflow:

1. Input: A nucleotide assembly (genome, transcriptome, metagenome).
2. Sequence Retrieval: Uses BLAST+ against a curated reference dataset derived from the Protist Ribosomal Reference (PR2) database. The reference covers nine major eukaryotic supergroups (e.g., Amoebozoa, TSAR, Archaeplastida).
3. Phylogenetic Placement: Retrieved SSU sequences are aligned (using MAFFT) and placed onto reference trees using pplacer (Maximum Likelihood mode). This allows for taxonomic assignment based on evolutionary position rather than simple sequence similarity.
4. Chimeric Detection: Uses VSEARCH (--uchime_ref) to identify chimeric sequences that may result from assembly errors or PCR artifacts.
5. Bacterial Contamination Proxy: Performs BUSCO searches against the bacterial bacteria_odb12 dataset. High scores indicate potential bacterial contamination (though the authors note this could also represent Horizontal Gene Transfer).
6. Output: Generates structured reports, summary statistics, and annotated phylogenetic trees (using ETE3) highlighting target sequences, contaminants, and chimeras.

Validation Case Study:
To validate the tool, the authors applied CSI-SSU to 2,960 genomic assemblies from the P10K database. They specifically focused on the Amoebozoa supergroup, performing independent phylogenetic reconstructions using:

• SSU (18S) rRNA for broad amoebozoan taxonomy.
• COI (Cytochrome c oxidase subunit I) for the testate amoeba order Arcellinida.
  These reconstructions utilized IQ-TREE with ModelFinder and ultrafast bootstrapping to verify CSI-SSU classifications.

3. Key Contributions

• Development of CSI-SSU: A scalable, reproducible, and modular tool (available on GitHub) that integrates sequence retrieval, chimera detection, and phylogenetic placement into a single workflow.
• Curated Reference Datasets: The creation of high-quality, non-redundant SSU reference alignments and trees for nine major eukaryotic supergroups, derived from PR2.
• Phylogenetic Validation: Demonstrating that phylogenetic placement is superior to BLAST for distinguishing between closely related taxa and identifying cross-group contamination.
• Comprehensive Benchmarking: A large-scale assessment of the P10K database, providing a "quality control" framework for future users of protist genomic data.

4. Key Results

• Efficiency: CSI-SSU processed 2,960 assemblies efficiently. Median runtimes were between 3.9 and 9.2 minutes, with memory usage typically under 500 MB, though large assemblies required up to 19.9 GB.
• Contamination Prevalence:
  • CSI-SSU retrieved 13,513 SSU sequences from the 2,960 assemblies.
  • Amoebozoa showed the highest levels of non-target contamination, while Obazoa (containing model organisms) showed the lowest.
  • TSAR (Stramenopiles, Alveolata, Rhizaria) was the most frequent source of cross-group contamination, followed by Obazoa and Archaeplastida.
  • Chimeras: Numerous chimeric sequences were detected across all supergroups.
• Taxonomic Corrections:
  • Phylogenetic analysis of Amoebozoa confirmed most original P10K assignments but corrected seven major misidentifications, particularly within the Arcellinida (testate amoebae).
  • Misidentifications often involved genera with convergent shell morphologies (e.g., Difflugia vs. Hyalosphenia).
  • The tool successfully refined genus- and family-level assignments for 43 taxa that were previously only identified at higher taxonomic levels.
• Bacterial Contamination: BUSCO searches revealed that many assemblies (especially in Amoebozoa and TSAR) contained significant numbers of bacterial markers (some approaching near-complete bacterial gene representation), suggesting incomplete decontamination in the original P10K pipeline.
• Quality Assessment: The study established criteria for "high-quality" assemblies: presence of the target SSU marker, no evidence of cross-group contamination, and a eukaryotic BUSCO completeness score ≥50%.

5. Significance

• Data Curation: CSI-SSU provides a critical "first-pass" filter for large-scale genomic databases, enabling researchers to distinguish high-quality datasets from those requiring decontamination or re-sequencing.
• Evolutionary Accuracy: By correcting taxonomic misidentifications and removing contaminants, the tool prevents artifactual phylogenetic relationships and erroneous functional annotations in downstream evolutionary studies.
• Scalability: The tool is designed to handle the growing volume of microbial eukaryotic data (e.g., P10K), offering a standardized framework for data validation that can be integrated into future sequencing pipelines.
• Future Directions: The authors emphasize that while CSI-SSU improves data reliability, future efforts must also focus on improving metadata documentation (e.g., voucher images, sequencing platforms) to fully realize the potential of microbial eukaryotic genomics.

In conclusion, CSI-SSU represents a significant step forward in ensuring the integrity of genomic resources for protists, transforming raw assembly data into reliable, phylogenetically validated datasets essential for understanding eukaryotic evolution.