Leviathan: A fast, memory-efficient, and scalable taxonomic and pathway profiler for (pan)genome-resolved metagenomics and metatranscriptomics

Leviathan is an open-source software package that enables ultra-fast, memory-efficient, and accurate taxonomic and functional profiling of metagenomes and metatranscriptomes at genome and pangenome resolution by combining alignment-free taxonomic methods with DNA-space pseudo-alignment to bypass computationally expensive translated-search steps.

The Gist

Imagine you have a massive library containing millions of books written by different authors (microbes), and you want to know two things: who wrote the books in a specific pile, and what stories (functions) those books tell.

For a long time, scientists trying to solve this puzzle used a method that was like reading every single word of every single book to find matches. It was incredibly accurate but painfully slow and required a supercomputer just to keep the lights on. This is the problem the paper addresses: existing tools were too slow and memory-hungry to handle the huge, modern collections of microbial "books" we now have.

Enter Leviathan, a new software tool designed to be the "express lane" for this kind of analysis. Here is how it works, using simple analogies:

1. The Speed Trick: Skipping the Reading

Instead of reading every word (which is what older tools did), Leviathan uses two clever shortcuts:

• The "Fingerprint" Scanner (Taxonomy): To figure out who is in the pile, it uses a tool called Sylph. Think of this like scanning a book's barcode or a unique fingerprint rather than reading the whole story. It instantly identifies the author without needing to read a single sentence.
• The "Table of Contents" Check (Function): To figure out what the microbes are doing, it uses a tool called Salmon. Instead of translating the text into a different language (a slow process called "translated-search" that older tools used), Leviathan looks at the "Table of Contents" (gene catalogs) directly in the original language. It matches the chapters it sees to the stories it knows, skipping the heavy translation step entirely.

2. The Double-Check System

Leviathan doesn't just guess; it gives you two specific scores for every story it finds:

• Abundance: "How many copies of this story are there?" (Like counting how many people are reading a specific book).
• Coverage: "Is the whole story there, or just a few pages?" It checks if the microbial community has all the necessary "chapters" (enzymatic steps) to complete a full metabolic pathway, ensuring the story makes sense from start to finish.

3. The Results: Faster and Lighter

When the authors tested Leviathan against the current gold standard (a tool called HUMAnN), the results were dramatic:

• Speed: It was up to 74 times faster. If the old tool took a week to finish a job, Leviathan could do it in a few hours.
• Memory: It used 14 times less computer memory. It's like running a marathon with a backpack full of bricks versus running with just a light jacket.
• Accuracy: It didn't just get faster; it got better at identifying the specific microbes and their genetic variations (pangenomes), improving accuracy by up to 12%.

4. Real-World Examples

The paper shows Leviathan in action with two specific stories:

• The Ocean Biofilm: They looked at microbes growing on plastic in the ocean. Leviathan helped them see how the "community conversation" changed as the biofilm grew from young to mature, revealing shifts in how they ate and survived.
• The Dental Caries Study: They analyzed the "voice" (gene activity) of bacteria in tooth decay. By looking at the specific genetic variations of the bacteria, they found unique patterns that distinguished between healthy mouths and those with cavities.

In short: Leviathan is a new, open-source tool that lets scientists analyze complex microbial communities much faster and with less computing power than before, without sacrificing accuracy. It's like upgrading from a slow, manual typewriter to a high-speed digital printer that also checks its own work.

Technical Summary

Technical Summary: Leviathan

Problem Statement
Functional profiling of metagenomes and metatranscriptomes is critical for deciphering the capabilities of microbial communities. However, existing methodologies face significant computational bottlenecks. Current approaches rely on computationally expensive translated-search alignments, which scale poorly against the rapidly expanding, large genome-resolved reference databases now standard in the field. This inefficiency limits the throughput and accessibility of high-resolution functional analysis.

Methodology
To address these limitations, the authors introduce Leviathan, an open-source software package designed for integrated taxonomic and functional profiling at both genome and pangenome resolutions. The core innovation lies in bypassing the translated-search step that dominates runtime in traditional pipelines. Leviathan achieves this through a two-stage hybrid architecture:

1. Taxonomic Profiling: It leverages Sylph to perform ultra-fast, alignment-free taxonomic classification.
2. Functional Quantification: It utilizes Salmon to conduct pseudo-alignment-based read quantification directly in DNA-space against genome-resolved gene catalogs.

For each (pan)genome, Leviathan generates dual functional metrics:

• Pathway Abundance: Derived from aggregated gene-level quantification.
• Pathway Coverage: Calculated via a graph-based assessment of enzymatic step completeness.

Key Results
The performance of Leviathan was evaluated against HUMAnN using the CAMI-I and CAMI-II benchmark datasets. The results demonstrate substantial improvements in efficiency and accuracy:

• Speed: Leviathan achieved runtimes up to 74-fold faster than HUMAnN.
• Memory: It demonstrated 14-fold lower memory usage.
• Accuracy: The tool improved genome-level assignment accuracy by up to 12% and pangenome-level accuracy by up to 5%.

Case Studies and Applications
The authors validated Leviathan's applicability through two specific case studies:

1. Marine Plastisphere Metagenomics: Differential coverage analysis revealed metabolic shifts between early and mature biofilm communities.
2. Dental Caries Metatranscriptomics: Pangenome-resolved co-expression network analysis identified organism-specific transcriptional patterns that serve as diagnostics for health and disease states.

Significance
Leviathan offers a scalable, memory-efficient solution for the modern era of (pan)genome-resolved metagenomics and metatranscriptomics. By eliminating the translated-search bottleneck, it enables high-throughput functional profiling that maintains or improves accuracy over existing standards. The tool is available as open-source software at https://github.com/jolespin/leviathan.