PREreview of A genomic history of African cattle

I wrote this review report for the Howard Hughes Medical Institute course Transparent and Accountable Peer Review with the support of my research mentor. My expertise lies in population genetics applied specifically to both ancient and modern human populations.

This study focuses on cattle genetic diversity in Africa and how that diversity has been affected by different environments and human interference. It analyzed ~4,900 cattle genomes, including 1,392 newly sequenced modern and two ancient genomes from Zimbabwe and South Africa from the past millennium. The authors used standard population genetics tools, such as PCA (principal component analysis), genetic clustering analysis, f-statistics, and local ancestry inference. They correlated climatic variables with ancestry distribution in different regions and did a selection scan with an IBD-based (identity by descent) tool. They found that 1) East African populations have more indicine ancestry than their Western counterparts, 2) taurine ancestry is geographically structured, 3) Southern African cattle groups are more closely related to East African ones, 4) African taurine ancestry is more strongly predicted by annual precipitation, while indicine ancestry was better predicted by temperature seasonality, and 5) a region originating in European taurines seems to be under selection in Southern African cattle. The claims are generally supported by the evidence presented, except for those based on the selection analysis, about which I have outlined recommendations below. The genetic dataset is of interest to the community, particularly the ancient genomes. This large and diverse dataset may be particularly useful for future investigations of finer structure in cattle genetic diversity in Africa.

Major comments

1) Lines 264-269: “[...] Both ancient and modern Southern African cattle were substantially closer to East African populations for both ancestry components”. For the ancient samples from Great Zimbabwe and Elsenburg (Extended Data Figure 1), the outgroup-f3 values for the taurine component vary little between comparison samples and there are plenty of samples in East Africa that share as little genetic drift as populations in the West (more visible in panel a). These statistics may be overlapping between comparisons when considering confidence intervals. The statement that Southern African samples are more closely related to East African samples requires further investigation. Firstly, for both indicine and taurine components, I recommend adding in the Supplementary Information File (SI) plots depicting along one of the axes the outgroup-f3 statistics point estimates and corresponding 95% interval as error bars and comparison populations along the other axis. Secondly, I recommend using haplotype-based methodologies such as Chromopainter and FineStructure to investigate the genomic history of Southern African cattle. Such analysis would provide the community with a more precise history of these cattle and it would make this study more impactful and novel.

2) The selection scan with isweep identified 22 genomic regions enriched for IBD and 21 of these were more commonly found in an African taurine background. Lines 377-378 highlight the only European taurine-specific region found to be under selection, overlapping with the PLAG1 gene. As isweep was developed on human genetic data, I am concerned that applying it with default parameters to cattle genetic data does not yield accurate results. I have listed my concerns and comments below:

• The first step of isweep is IBD detection using hap-ibd and ibd-ends. In my experience working with human data, IBD detection is challenging and it depends on phasing accuracy. I recommend benchmarking IBD detection for cattle data using the methods that are part of the isweep workflow, to determine whether IBD sharing in cattle can be accurately detected as part of this analysis.

• In isweep, it is assumed that the samples are unrelated and this is a panmitic population. In the methods section, I could not find whether kinship was estimated and whether the input dataset was split by breed or by a grouping that reflects closely shared ancestry in contrast to using the full dataset. Different breeds have different demographic histories, including different effective population sizes (Ne), important for the IBD simulation step. Moreover, there are likely different selection processes via human intervention. As such, the assumption of a panmitic population is not met. If indeed the full dataset was used, I suggest splitting the dataset into groups and use isweep on each group and then compare those results with what the study has already found. Sample size per group may be an issue, though. Therefore, and based on the general comment that this tool was calibrated on human data, I recommend simulating cattle-like genetic data to investigate whether isweep is an appropriate tool to detect selection on cattle.

• No investigation into whether each candidate region is a true or false positive has been carried out. Multiple phenomena may lead to an excess of IBD sharing in a region, such as low recombination rate, low SNP density, or location within an inversion. I suggest cross-referencing these features with the candidate regions. This would allow identification of regions that are more likely to be under selection.

• A single methodology has been used to detect local adaptation. I suggest using other complementary approaches, such as extended haplotype homozygosity (EHH) and detection of regions enriched for a specific ancestry (African taurine vs. indicine vs. European taurine). Using multiple methodologies may strengthen the conclusions of this analysis and help assess how robust each candidate region is.

• To make this section of the manuscript more complete, I would recommend listing all the genomic regions of interest in a supplementary table and not only the European taurine region, including their estimated selection coefficients and whether these overlap functional regions, in which case what these are and what is known about them.

• Previous studies found regions under selection in African cattle (e.g., Kim et al. 2020 https://doi.org/10.1038/s41588-020-0694-2, McHugo et al. 2024 https://doi.org/10.1038/s41437-024-00734-w, and Gao et al. 2026 https://doi.org/10.1038/s42003-026-09856-9). The conclusions of this analysis would be strengthened by putting these results into the context of previous studies that used different datasets and methodologies.

• Figure 3b depicts haplotypic frequency for the detected region overlapping PLAG1 in different African cattle. Is it known whether this haplotype is also at a high frequency in European cattle? I suggest including European breeds and estimate the frequency for this region for those, as it would provide extra information on whether this is a region that has been under selection in other domestication contexts as well.

Minor comments

1) Lines 161-165: “All four populations associated with the pastoralist groups, White Fulani, Sokoto Gudali, Mbororo, and Zebu Peul, exhibit a relatively uniform (~63%) indicine admixture proportion. These near-identical values suggest that these cattle populations have a common origin, [...]”. Concluding that populations have a common origin based on having similar admixture proportions is an overstatement, particularly when those admixture proportions are estimated using a genetic clustering tool such as ADMIXTURE More formal tests could be done, using approaches such as qpWave, qpAdm, or using haplotype-based methods, such as detecting IBD (identity by descent) sharing between these populations and others to examine whether they have recent genealogical connections. Such tests would provide us with a more complete view of the demographic history of these cattle populations.

2) Lines 168-171: “A second, more recent contribution to indicine ancestry in West Africa may be linked to the Great African Rinderpest Pandemic of 1887-97, which killed an estimated 80-90% of cattle across much of sub-Saharan Africa and triggered extensive restocking from surviving herds.” An interesting analysis for this study would be estimating effective population size changes across different breeds, particularly for more recent time periods as the one mentioned above. This would provide the cattle genetics community with a more precise demographic history of cattle in Africa.

3) Throughout this manuscript, I missed information that is relevant for readers to interpret the results and for reproducibility of the study itself. I suggest re-organizing the SI file into Supplementary Notes that can be referenced in the main text. The SI has information that is necessary to interpret the methods and figures. I recommend making the main text more complete so that readers can evaluate the study and potentially reproduce it. I am listing below some examples of information that is missing:

• Assumed generation time: Figure 1d and dating of admixture (methods)

• Outgroup used for f-statistics: Extended Data Figures 1 and 2 and methods

• Confidence intervals in Figure 1d (95%?)

• What are the genome-wide significance thresholds used in Figure 3?

• Number of samples used for each ancestry component in Extended Data Figure 1

• Supplementary Figure 12: What are the D/F4-statistic tests? What do the colors represent? What do the error bars indicate?

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.