Abstract Comparative oncology studies, for example those encompassing non-human mammals, offer unique opportunities to understand human cancer susceptibility and resistance. A major challenge hindering these studies is inconsistent and incomplete metadata. To address this challenge, we developed Paipu, a computational pipeline that efficiently and systematically retrieves and harmonizes RNA-seq sample data. Paipu consists of three computational phases: reference genome preparation, SRA metadata retrieval and harmonization, and RNA-seq data processing and harmonization. In the genome preparation phase, Paipu retrieves the highest quality reference files for each species, builds genome indexes, and prepares annotations and species-specific configuration files. In the SRA metadata retrieval phase, Paipu addresses challenges such as misspellings, mislabeling, and other data inconsistencies. The RNA-seq data processing phase uses FREYA to generate high-quality gene expression counts for each sample. To demonstrate Paipu’s effectiveness, we created a pan-mammalian pan-cancer dataset. We applied Paipu to the 240 mammalian species (excluding humans) in the Zoonomia alignment and 188 cancer-related search terms, identifying 135, 713 unique samples spanning 21 species. After Paipu removed duplicates and low-quality samples, the dataset contained 121, 786 RNA-seq samples from 21 of 240 non-human mammalian species. Preliminary analyses of these data show that most samples are from mice (96%), dogs (2%) and rats (1%). The majority of samples are paired-end reads, and the dataset contains ∼56, 000 bulk and ∼65, 000 single-cell samples. We analyzed key metadata attributes including cancer type, species, breed and single or bulk sequencing. We grouped cancer types by organ system, reducing 83 cancer types to 13 categories. Prior to data harmonization, we identified 13 metadata columns required by NCBI SRA. Despite this, on average only 81% of samples had this required information, as identified by Paipu. An additional 1, 407 other columns are not required by SRA but were included in at least one dataset, largely due to the variety of experimental designs for SRA studies. Consequently, many columns are only defined for a small percentage of the data. This results in an average of ∼96% missing values in the integrated dataset. Paipu harmonizes these columns when possible. For example, Paipu identified ∼20 columns specifying organ or tissue type and ∼4 columns specifying sample sex, which it reduced to a single column for each. Our results demonstrate that effective metadata harmonization is crucial for enabling pan-mammalian comparative oncology studies. By revealing patterns of experiments and species-specific representation, Paipu facilitates identification of suitable model organisms for rare human cancers and supports large-scale analyses. Citation Format: Bria S. Smith, Leslie A. Smith, James A. Cahill, Kiley Graim. The Paipu framework enables large-scale comparative cancer genomics studies abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Cancer Evolution: The Dynamics of Progression and Persistence; 2025 Dec 4-6; Albuquerque, NM. Philadelphia (PA): AACR; Cancer Res 2025;85 (23Suppl): Abstract nr B040.
Leslie Smith (Thu,) studied this question.