The following archives contain data files associated with the manuscript:

Kathryn A. Lord, Vista Sohrab, Kasia Bryc, Michelle E. White, Brittney Kenney, Kathleen Morrill Pirovich, Frances L. Chen, Elinor K. Karlsson, Genetic testing predicts appearance but not behavior in dogs. PNAS. 2025.

Data Files

Genetic Data:

darwins_dogs_genetic_set.zip Contains a subset of the gene dataset.

Darwin's Ark Cohort

A PLINK1 bfile set (.bed/.bim/.fam) with sample IDs encoded as dog IDs matching individuals from the Darwin's Ark survey data files, including genotypes from 3,277 dog whole genomes sequenced under BioProject PRJNA675863.

DarwinsDogs_2024_N-3277_canfam4_gp-0.70_biallelic.bed

DarwinsDogs_2024_N-3277_canfam4_gp-0.70_biallelic.bim

DarwinsDogs_2024_N-3277_canfam4_gp-0.70_biallelic.fam

The following plink set above is then further filtered using minor allele frequency and Hardy-Weinberg p-value thresholds for genetic analysis. See the Methods section of the manuscript for more details.

Genetic analysis input files

darwins_dogs_gwas_input_files.zip contains the phenotype input files and covariate input files used to run the analysis.

phenotype_input_files

This folder contains the following files:

• DarwinsArk_4Q_Morphology_N-3277_rerun_20241107.tsv Contains level phenotypic data for morphological questions that were recoded differently from the raw survey questions.
• DarwinsArk_13Q_Q243_coat_color_N-1930.tsv Contains level-phenotypic data for the coat color question.
• DarwinsArk_34Q_Priorities1-3_Behavior_9Q_Morphology_N-3277.tsv Contains level-phenotypic data for behavioral questions and morphological questions (raw data in this file and no recoding for morphological questions).
• DarwinsArk_8Factors_N-3277.tsv Contains I-level behavioral factor scores.

Each phenotype input file requires a required dog_id column referring to the dog_id used in both our phenotypic and genetic datasets, followed y columns for the phenotype of interest.
The workflow used to run our genetic analysis creates an individual phenotype file from the following input files above.
Workflow for analysis can be found at https://github.com/VistaSohrab/dog-gwas-heritability-nextflow

FDarwinrwi.n's Ark questio.n key from abbreviation to full text (ie, Q121 refers to dog height), see Supplemental Table 5 (Table S5) of the manuscript.

covariate_input_files

This folder contains the following files:

• DarwinsArk_AgeComposed_Height_N-3277.tsv contains individual-level ages and heights used in all GWAS analyses as a quantitative covariate, except for height GWAS
• DarwinsArk_AgeComposed_N-3277.tsv contains individual-level ages used as a quantitative covariate in height GWAS
• DarwinsArk_Sex_NeuterStatus_N-3277.tsv contains individual-level sex status (sex and neuter status) as a single discrete covariate used in all the GWAS

Genetic analysis output files (GWAS)

darwins_dogs_gwas_output_files.zip contains all GCTA leave-one-chromosome-out mixed linear model association output files (.loco.mlma)

This folder contains GWAS output files for behavioral and morphological questions, as well as behavioral factors.

Heritability output files (GCTA-GREML)

darwins_dogs_heritability_output_files.zip contains heritability output files from GCTA-GREML

This folder contains GCTA-GREML output files (both LD-corrected and non-LD-corrected) (both constrained and unconstrained)

• .lds.hsq refers to LD-corrected and constrained (most reliable and reported in the manuscript)
• .lds.no-constraint.hsq refers to LD-corrected and no constraint
• .no-lds.hsq refers to non-LD-corrected and constrained (can refer to this file in case the LD-corrected and constrained run was unsuccessful)
• .no-lds.no-constraint.hsq refers to non-LD-corrected and no constraint

Supplemental information: GWAS analysis

supplemental_information_gwas.zip contains GCTA leave-one-chromosome-out mixed linear model association output files (.loco.mlma) from 5 phenotypes (2 behavioral questions, 2 behavioral factors, and height as morphological question) pera formed on mixed breed random sampling of mixed-breed and single-breed dogs from the full dataset.

• random_dogs In the GWAS output file name or the phenotype input file column name refers to dogs randomly sampled from the full dataset
• mixed_dogs In the GWAS output file or the phenotype input file column name refers to mixed-breed dogs

The phenotype input files can be found in the supp_gwas_phenotype_input_files folder. Covariate files are the same as other GWAS (age and height, sex status) or (age, sex status) for height.

Data availability

FASTQ files for all genetic information in this study are available in the NIH Sequence Read Archive (project PRJNA675863) at https://www.ncbi.nlm.nih.gov/sra?term=PRJNA675863. The genetic data, as well a,s the GWAS and heritability input and output files are avail, are available on Dryad at (DOI: 10.5061/dryad.83bk3jb4r). The code for curating the phenotypes and predicting SNP effects on phenotype is available on GitHub at https://github.com/broadinstitute/dog_behavioral_gwas_paper. The pipeline for GWAS and heritability is available at https://github.com/VistaSohrab/dog-gwas-heritability-nextflow/ with the necessary software packaged in a Docker container at https://hub.docker.com/r/vistasohrab/terra-dogagingproject-gwas.

Contact

If you have further questions about the data file archive, then contact the authors, Kathryn L. Ord (kathryn.lord@umassmed.edu) and Elinor Karlsson (elinor.karlsson@umassmed.edu), and we will resolve them as soon as possible.