Genome-wide selection scans

[hammer]

From @alimanfoo :

Detecting signals of recent positive selection is likely to be a major focus for MalariaGEN, because this is how we will detect the emergence of new forms of drug and insecticide resistance as new drugs and insecticides are deployed in the field, and in response to other changes happening in disease and vector control intervention programmes.

Note that these selection scans do not need to scale to massive numbers of samples, because usually you are running the scan for one population sample at a time, and you generally don't need more than ~100 individuals in any given population sample to run the scan. So the goal here is to be able to run these at interactive speed, for multiple population samples. I.e., we're working towards sampling mosquitoes from up to ~100 locations, and at least twice per year at each location, so you want to run these scans for each location and time point to see when and where new selection signals are occurring.

If you want to get a sense of what these selection scans look like in real data, have a browse around the Ag1000G selection atlas. E.g., here's a stonking signal of selection over an insecticide resistance gene cluster.

Population branching statistic (PBS)

• Runs on unphased data.
• Notebook from Alistair illustrating its use on Ag1000G phase 2 data.
• Sequencing of Fifty Human Exomes Reveals Adaptation to High Altitude (2010) first proposed PBS. The supplementary material compares PBS to Tajima’s D, and says that PBS

is similar to the “locus-specific branch lengths” statistic used by The genomic distribution of population substructure in four populations using 8,525 autosomal SNPs (2004), except that by using the above log-transformation, we obtain additive distances that place branches of different magnitudes on the same scale.

• PBS uses population trios and computes fixation index (F_ST) distances to compare two populations of interest to an anchor population.
• Luca Pagani's doctoral thesis, Through the layers of the Ethiopian genome: a survey of human genetic variation based on genome-wide genotyping and re-sequencing data (2013), contains an explanation of PBS with a diagram.
• Paula Grubb's master's thesis, Evolution of Malaria Resistance in Africa and Island Melanesia (2016), also includes an explanation of PBS with diagrams.

Haplotype homozygosity (H12)

• Computed from haplotype (phased) data.
• moving_garud_h from scikit-allel.
• Selection scans blog post from Nandita Garud, first author on the below papers. Code for this blog post is at ngarud/SelectionHapStats.
• Recent selective sweeps in North American Drosophila melanogaster show signatures of soft sweeps (2015) proposes H12 and H2/H1.
• Enhancing the mathematical properties of new haplotype homozygosity statistics for the detection of selective sweeps (2015) improves them.
• Detection and Classification of Hard and Soft Sweeps from Unphased Genotypes by Multilocus Genotype Identity (2018) defines G12, G2/G1, and G123, which can be computed for multi-locus genotype data without phasing/haplotype information!

Other

• Detecting recent positive selection in the human genome from haplotype structure (2002): Long-Range Haplotype (LRH)
• A map of recent positive selection in the human genome (2006): integrated Haplotype Score (iHS)
• Genome-wide detection and characterization of positive selection in human populations (2007): Cross Population Extended Haplotype Homozogysity (XP-EHH)