Recomb-Mix Cloud is a cloud-friendly, end-to-end pipeline developed to perform local ancestry inference (LAI) on large genomic datasets typically hosted in cloud environments. The pipeline is built on the Recomb-Mix LAI method, which evaluates ancestry at the level of individual variants and is designed to reduce computational burden relative to window-based LAI approaches. The project aims to make LAI feasible and affordable for researchers working on biobank-scale data, particularly those with limited computing resources.
Recomb-Mix Cloud delivers a complete workflow for LAI on cloud-hosted data. The pipeline includes the following stages: panel sample selection, marker alignment, construction of a compact reference panel, Recomb-Mix threading, and generation of LAI calls in standard data formats. By packaging these steps into a cloud-ready pipeline, the authors intend to simplify the practical steps required to run LAI from cloud platforms where many large genomic resources reside.
The core LAI engine is Recomb-Mix, which focuses on individual genetic variants rather than genomic windows. According to the authors, this design maintains accuracy while reducing computational demands compared with methods that operate on windows of markers.
To support immediate use, the authors constructed a set of pre-computed compact population reference panels. These panels were derived from public datasets including the Human Genome Diversity Project and the 1000 Genomes Project. The pre-computed panels are distributed with the pipeline, reducing the need for users to build large reference resources themselves.
The Recomb-Mix Cloud pipeline and the compact reference panels are available from the project repository: https://github.com/ucfcbb/Recomb-Mix. Users can employ the provided panels or create their own compact panels using the pipeline’s panel construction steps.
The authors compared Recomb-Mix (as implemented in Recomb-Mix Cloud) to two widely used LAI tools, FLARE and RFMix, on the All of Us Researcher Workbench. They report that Recomb-Mix is approximately one to two orders of magnitude cheaper and faster than these tools while producing largely consistent LAI calls on the same data platform.
A key metric reported is the estimated overall compute cost per All of Us whole-genome sample when using a pre-computed reference panel: $0.01 per sample for Recomb-Mix Cloud, compared with $0.43 for FLARE and $4.58 for RFMix. These cost estimates illustrate the potential for substantial savings when performing LAI at scale in cloud environments.
Recomb-Mix Cloud is explicitly designed to operate in cloud-hosted settings where many modern biobank datasets reside. By minimizing computational resource requirements and providing compact reference panels, the pipeline is positioned to scale across large cohorts with lower operational cost. The reported results on the All of Us Researcher Workbench demonstrate the pipeline’s application on a major cloud-hosted research resource.
The authors emphasize that the combination of variant-level LAI, compact reference panels, and a cloud-optimized workflow supports democratizing ancestry-based analyses for researchers who may not have access to large local compute clusters.
The pipeline source code and compact reference panels are publicly available at the project GitHub: https://github.com/ucfcbb/Recomb-Mix. The distribution of pre-computed panels based on the Human Genome Diversity Project and the 1000 Genomes Project is intended to facilitate reproducible analyses and to lower setup time for new users.
The manuscript notes NIH funding support. The authors declared no competing interests.
This work is presented as a preprint on bioRxiv and has not been certified by peer review. Users should interpret results with that context in mind. Detailed implementation choices, benchmarking protocols beyond the summary results reported here, and further validation specifics are contained in the full preprint; if additional methodological or benchmarking details are required, readers should consult the full text and supplementary material at the repository and the bioRxiv entry.
Recomb-Mix Cloud packages the variant-focused Recomb-Mix LAI method into a cloud-optimized pipeline that includes sample selection, marker alignment, compact reference panel construction, threading, and output in common formats. The authors report that, when applied on the All of Us Researcher Workbench, Recomb-Mix Cloud achieves markedly lower compute cost per sample (reported as $0.01) and faster runtimes compared with FLARE and RFMix while producing largely consistent LAI calls. Pre-computed compact reference panels are provided to simplify adoption, and the pipeline is available on GitHub for researchers to deploy on cloud platforms. The manuscript is a preprint and has not undergone peer review.