Population genomics in practice

What is population genomics?

Per Unneberg

Intended learning outcomes

Course

• Present minimum toolkit of methods that should be known to anyone starting out in population genomics
• Sufficiently small for one-week workshop

Lecture

• Present practical example of toolkit as applied in (Fuller et al., 2020)
• Briefly discuss baseline model (Johri et al., 2022)

Aim of lecture is to:

• present a practical application of commonly used methods in population genomics
• link population genomics to population genetics
• discuss statistical inference and the need of a baseline model with which to compare observations and conclusions

What is population genomics?

Points from (Hahn, 2019, pp. 249–250):

• whole-genome data instead of single loci - population genomics is population genetics for whole-genome sequences
  • if only this, not too exciting
• major promise: enables analyses not possible for single loci or that require genomic context
• addresses interactions between different forces, notably selection and demographic history

Some applications

• genome-wide scans for selection
  • selection vs demography (p. 251)
• methods for genome-wide scans (p. 258)

Caveats

• non-independence (p. 267)
  • different statistics rely on similar input
  • overlapping peaks from different statistics not independent

General points

(Hartl & Clark, 1997, pp. 469–470):

• more emphasis on differences within populations
• goal: understand differences among genomes -> requires complete sequence data from multiple individuals

(Li & Durbin, 2011, supplementary notes, p. 6) on the use of PSMC on autosomes:

“…highly consistent except for the very recent history, demonstrating the power of using whole-genome data.”

Example: Population genetics of the coral Acropora millepora

Motivation: corals are facing hard times and to prevent future losses of coral cover a better understanding of genetics is warranted.

Genome assembly and sampling

Motivation: most analyses require a reference sequence with which to compare resequenced samples

• Assemble high-quality reference genome
• Choice of populations, sampling locations

Figure 1: Genome assembly by hybrid sequencing (A) 253 individuals from 12 reefs (B) PCA with environmental and spatial variables (C) phenotype distributions (D)

Fuller et al. (2020)

(Fuller et al., 2020) is an example of a population genomics study that applies methods that could be seen as a basic foundation of population genomics. We believe these present a minimum toolkit of methods that should be known to anyone starting out in population genomics, and that is sufficiently small to be presented in a one-week workshop. At the end of this lecture, we will discuss some more advanced applications in population genomics.

Figure caption:

Genome assembly and sample collection for A. millepora. (A) A de novo assembly for the A. millepora genome was constructed by using a hybrid sequencing approach. The alignment of reads generated from two pools of aposymbiotic larvae was used to filter out symbiont contigs (fig. S3). Assembled coral contigs were aligned to previously published linkage maps (19, 80) to create a chromosome-scale assembly. (B) A total of 253 individuals were collected from across 12 reefs on the GBR in 2017. The size of the circles represents the number of individuals collected at each site, and the circles are colored according to the bleaching severity at each reef. The maximum degree heating weeks (DHW) is shown across the region from which samples were collected (and across the GBR in the inset). The spatial layer of DHW represents interpolated maximum values from the National Oceanic and Atmospheric Administration Coral Reef Watch (CRW) v3.1 satellite product at a resolution of 5 km. Each reef label is colored arbitrarily but consistent with labels presented in other figures. AN, Arlington Reef; FY, Fitzroy Reef; RL, Russell Island; CS, Coates Reef; FR, Feather Reef; NB, North Barnard Islands; TR, Taylor Reef; DK, Dunk Island; RB, Rib Reef; PA, Pandora Reef; JB, John Brewer Reef; HH, Havannah Island. (C) A PCA was performed for 40 environmental and spatial variables for each reef, with abbreviations in bold representing their location. The component loads for each environmental variable projected on the first two principal components are depicted with arrows. (D) The distribution of visual scores, chlorophyll abundance standardized by host coral protein content, and standardized symbiont cell densities among the collected individuals for which phenotype measurement was possible. A visual score of 1 indicates severe bleaching, whereas a score of 6 represents full pigmentation.at a resolution of 5 km. Each reef label is colored arbitrarily but consistent with labels presented in other figures. AN, Arlington Reef; FY, Fitzroy Reef; RL, Russell Island; CS, Coates Reef; FR, Feather Reef; NB, North Barnard Islands; TR, Taylor Reef; DK, Dunk Island; RB, Rib Reef; PA, Pandora Reef; JB, John Brewer Reef; HH, Havannah Island. (C) A PCA was performed for 40 environmental and spatial variables for each reef, with abbreviations in bold representing their location. The component loads for each environmental variable projected on the first two principal components are depicted with arrows. (D) The distribution of visual scores, chlorophyll abundance standardized by host coral protein content, and standardized symbiont cell densities among the collected individuals for which phenotype measurement was possible. A visual score of 1 indicates severe bleaching, whereas a score of 6 represents full pigmentation.

Genome assembly and sampling

Why: most analyses require a reference sequence with which to compare resequenced samples

Points to consider:

• choice of reference individual
• the number of populations
• the number of samples (more sites better than many samples per population)
• the geographical distribution of samples
• sequencing depth (low-coverage often sufficient)

Example: Population genetics of the coral Acropora millepora

Motivation: corals are facing hard times and to prevent future losses of coral cover a better understanding of genetics is warranted.

Describe genetic structure and demographic history

Motivation:

• address basic question of why genetic structure looks the way it does
• demographic history may generate signals similar to selection

Figure 2: Variation and demographic history inferred from 44 resequenced individuals. LD decay (r^2) from 1% of markers (A) nucleotide diversity (\pi) in 1kb-windows (B) effective population sizes (\mathrm{N_e}) estimated with PSMC (C)

Fuller et al. (2020)

Figure caption:

Properties of genetic variation and inferred demographic history in sampled A. millepora. (A) The decay of LD, measured as the squared genotypic correlation coefficient (r2), was estimated for a randomly chosen 1% of SNPs by using all sequenced samples. Each point represents the mean r2 in bins of 100 bp. (B) Nucleotide diversity was measured by the average pairwise differences per base pair (p) (86) for SNPs in 1-kb windows, using intergenic regions genome-wide. (C) Effective population sizes (Ne) inferred from 44 sequenced A. millepora genomes by using PSMC (35), assuming a mutation rate m of 4 × 10−9 per base pair per generation (20, 34). Although an error in the estimate of m will lead to a rescaling of both axes, it will not affect the qualitative conclusion of a decline in Ne toward the present. The x axis is scaled in terms of generations. This approach provides very little information about the recent past (here, less than ~104 generations).

Variation and demographic history

Why: summarizing diversity provides (indirect) information on population size and more, as does the linkage structure. Estimate demographic history since fluctuating population size may produce signals similar to those of selection

• LD decay: important for imputation (e.g., stephens_AccountingDecayLinkage_2005) and setting window size for genome scans, where a common rule of thumb is to set the size larger than the genome background: this ensures windows are, in some sense, independent
• “The extent of LD and its decay with genetic distance are useful parameters for determining the number of markers needed to successfully map a QTL, and the resolution with which the trait can be successfully mapped” [otyama_EvaluationLinkageDisequilibrium_2019]
• 0.363% average pi, but large variation.
• many psmc plots show decline in population size, which could be an effect of bottleneck during pleistocene. Also population divergence (ghost ancestral populations, splits, extinction) can affect population size
• in aDNA studies missingness is common (i.e., heterozygotes are underestimated) and has to be accounted for since coalescence times are affected and may influence estimate of population size

Example: Population genetics of the coral Acropora millepora

Motivation: corals are facing hard times and to prevent future losses of coral cover a better understanding of genetics is warranted.

Characterize population structure

Motivation:

1. identify populations for contrasts in e.g. selection scans
2. identify admixed individuals that should be removed from analyses
3. identify barriers to gene flow etc

Figure 3: Characterizing population structure and gene flow across 12 refs. F_{\mathrm{ST}} measure across geographic distance (A) PCA from LD-pruned genome-wide SNPs for 44 resequenced samples (B) estimation of relative effective migration surface from LD-pruned and common (MAF > 5%) SNPs (C)

Fuller et al. (2020)

Characterizing population structure and gene flow across 12 reefs. (A) There is no discernable relationship between geographic distance and genetic differentiation (measured as FST) across pairwise comparisons of sampled reefs. (B) A PCA by using LD-pruned genome-wide SNPs for the 44 resequenced high-coverage genomes. Each point is color-coded by the reef from which the individual was sampled. (C) EEMS (39) was used to estimate and visualize the relative effective migration surface by using LD-pruned and common (MAF > 5%) genome-wide SNPs for the 44 resequenced high-coverage genomes. The size of the dots reflects the number of individuals sequenced from each reef.

Population structure:

Why: many reasons: 1) identifying populations for contrasts in e.g. selection scans 2) identify admixed individuals that should be removed from analyses 3) identify barriers to gene flow etc

• no discernible relationship between geographic distance and genetic differentiation -> gene flow
  • for this reason, Fst between populations is low
• EEMS (Estimated Effective Migration Surfaces) models relationship between genetics and geography petkova_VisualizingSpatialPopulation_2016

Indicative of high connectivity among 12 sampled reefs.

Example: Population genetics of the coral Acropora millepora

Motivation: corals are facing hard times and to prevent future losses of coral cover a better understanding of genetics is warranted.

Genomic scans for selection

Motivation: identify loci associated with adaptation / selection

• little differentiation over reefs, however thermal regimes
• genomic scan for \pi (diversity) outliers

Figure 4: Genomic scans for local adaptation detect a signal at sacsin. Nucleotide diversity (\pi) in 1kb-window, values in top 0.01% genome-wide in red (A, top). Close-up view around sacsin gene, with predicted gene structure above (A, bottom). h12 summary statistic showing the frequence of the two most common haplotypes (B). sacsin gene tree (blue) together with random gene trees, indicating deep genealogy at sacsin (C)

Fuller et al. (2020)

Genomic scans for local adaptation detect a signal at sacsin. (A) (Top) Values of pairwise nucleotide diversity (π) (86) estimated in sliding windows of 1 kb and a step-size of 500 bp across chromosome 7, with points colored in red indicating regions in the top 0.01% of values genome-wide. A loess-smoothed trend line of π across windows is indicated in purple. The peak with the most extreme values falls in the sacsin gene region. (Bottom) A close-up view of π estimated across a region of chromosome 7 surrounding the sacsin gene, which includes all of the outlier windows indicated in red at top. π per base pair was calculated in sliding windows of 100 bp (gray dots); the loess-smoothed line is shown in blue. The predicted gene structure of sacsin is indicated above, as well as the flanking upstream and downstream genes. The 29 nonsynonymous differences fixed between samples of the two most common haplotypes in sacsin are denoted with green lines. There are also 20 fixed synonymous differences. Gray vertical dashed lines delimit a region that was masked from variant calling because of predicted repetitive elements. (B) The h12 summary statistic (47), which measures the frequency of the two most common haplotypes, is plotted across chromosome 7 (supplementary materials, materials and methods). Red dots represent h12 values in the top 0.01% genome-wide. The peak with the most extreme values falls in the sacsin gene region. (C) A gene tree for the central 1-kb region in sacsin is shown in dark blue; it was constructed for one randomly chosen haplotype from a randomly selected individual from each of the 12 sampled reefs. The gene tree is rooted to aligned sequences from the reference genomes for A. digitifera (87) and A. tenuis (draft assembly from <www.reefgenomics.org>). Shown in gray are gene trees for 1000 randomly sampled 1-kb regions from across the genome for the same individuals. Each gene tree was inferred by using maximum likelihood implemented in dnaml (88).

Selection scan

Why: identify loci associated with adaptation / selection, which provides potential mechanisms for adaptation, as well as information that could be important for conservation strategies

Little differentiation across reefs -> little population structure over hundreds of kilometers. However, there are environmental differences (thermal regimes). Scan for pi outliers:

• points to sacsin gene
• h12 measures the frequency of the two most common haplotypes; red indicate 0.01% outlier genome-wide
• 4C: tree for central 1kb region in sacsin deeper than split from A.digitifera and A.tenuis
  • variation in sacsin has been maintained for long time
  • co-chaperone for heat-shock protein Hsp70

Example: Population genetics of the coral Acropora millepora

Study highlights common analyses in population genomics study:

1. Genome assembly, resequencing, variant calling and filtering
2. Description of variation (e.g., \pi) and genetic structure (LD)
3. Description of population structure (admixture, PCA)
4. Modelling of demographic history (PSMC)
5. Genome scans for adaptive traits

Population genetics

 

Mutation

Selection

 

Recombination

Drift

(Fuller et al., 2020) paper has population genetics in title -> population genetics is a key ingredient.

Population genetics focuses on the genetic basis of evolution. It is mainly a theoretical subject, owing to the slow changes of genetic variation. As such, it tries to explain the shape and structure of genetic variation from theoretical predictions and models.

From population genetics to population genomics

The variable sites at the Drosophila melanogaster ADH locus (Kreitman, 1983)

First study of natural population. However, limited to one locus.

• from locus-based studies (e.g., alcohol dehydrogenase in Drosophila (Kreitman, 1983)) to genome-wide (e.g., Drosophila population genomics (Begun et al., 2007)
• note: studied loci have often not been randomly chosen, which is another argument for whole-genome studies
• enabler: sequencing technology

(Fuller et al., 2020) paper has population genetics in title -> population genetics is a key ingredient.

Refer to Hahn’s points about learning something about global patterns:

• selection acts locally, demography globally
• the structure of genetic variation and how it depends on
  • recombination landscapes and linked selection
  • demographic changes
  • identification of neutral loci

So, not simply about applying 10000 selection tests for multiple loci

All of the points above point to the importance of statistics which implies mathematics / computational skills important

From population genetics to population genomics

Patterns of polymorphism and divergence (Begun et al., 2007)

Same system but genome-wide. Plots represent all chromosomes and the entire genome.

Begun et al. (2007) study: same system (Drosophila) but more individuals and whole genome. All of a sudden possible to ask questions about the general characteristics of diversity, not just limited to single loci.

From population genetics to population genomics

Numbers of polymorphic and fixed variants (Begun et al., 2007)

Novelty: now possible to do genome-wide characterization of variation in different functional contexts

Novelty: now possible to do genome-wide characterization of variation in different functional contexts

The technological revolution in sequencing and computing

Figure 5: Sequencing cost ($) per megabase (Wetterstrand, KA)

Moore’s law

Statistical inference in population genomics

The data deluge requires advanced statistical methods and models to do inference. Today data production outpaces theoretical advances. Therefore, take care not to attach too much faith to a test that explains data well.

A population genomics study should aim at generating a baseline model that takes into account the processes that shape genetic variation (Johri et al., 2022):

1. mutation
2. recombination
3. gene conversion
4. purifying selection acting on functional regions and its effects on linked variants (background selection)
5. genetic drift with demographic history and geographic structure

Caution against adaptationist storytelling; always compare to a baseline model that takes potential confounding factors into account

Applications of population genomics

Conservation genomics (Webster et al., 2023)

Speciation genomics (Stankowski et al., 2019)

disentangle forces that create variation (Rodrigues et al., 2024)

paleogenomics (aDNA) (van der Valk et al., 2021)

domestication (Barrera-Redondo et al., 2020)

ecology (Unneberg et al., 2024)

Bibliography

Barrera-Redondo, J., Piñero, D., & Eguiarte, L. E. (2020). Genomic, Transcriptomic and Epigenomic Tools to Study the Domestication of Plants and Animals: A Field Guide for Beginners. Frontiers in Genetics, 11.

Begun, D. J., Holloway, A. K., Stevens, K., Hillier, L. W., Poh, Y.-P., Hahn, M. W., Nista, P. M., Jones, C. D., Kern, A. D., Dewey, C. N., Pachter, L., Myers, E., & Langley, C. H. (2007). Population Genomics: Whole-Genome Analysis of Polymorphism and Divergence in Drosophila simulans. PLOS Biology, 5(11), e310. https://doi.org/10.1371/journal.pbio.0050310

Fuller, Z. L., Mocellin, V. J. L., Morris, L. A., Cantin, N., Shepherd, J., Sarre, L., Peng, J., Liao, Y., Pickrell, J., Andolfatto, P., Matz, M., Bay, L. K., & Przeworski, M. (2020). Population genetics of the coral Acropora millepora: Toward genomic prediction of bleaching. Science, 369(6501), eaba4674. https://doi.org/10.1126/science.aba4674

Hahn, M. (2019). Molecular Population Genetics (First). Oxford University Press.

Hartl, D. L., & Clark, A. G. (1997). Principles of population genetics. Sinauer Associates.

Johri, P., Aquadro, C. F., Beaumont, M., Charlesworth, B., Excoffier, L., Eyre-Walker, A., Keightley, P. D., Lynch, M., McVean, G., Payseur, B. A., Pfeifer, S. P., Stephan, W., & Jensen, J. D. (2022). Recommendations for improving statistical inference in population genomics. PLOS Biology, 20(5), e3001669. https://doi.org/10.1371/journal.pbio.3001669

Kreitman, M. (1983). Nucleotide polymorphism at the alcohol dehydrogenase locus of Drosophila melanogaster. Nature, 304(5925), 412. https://doi.org/10.1038/304412a0

Li, H., & Durbin, R. (2011). Inference of human population history from individual whole-genome sequences. Nature, 475(7357), 493–496. https://doi.org/10.1038/nature10231

Rodrigues, M. F., Kern, A. D., & Ralph, P. L. (2024). Shared evolutionary processes shape landscapes of genomic variation in the great apes. Genetics, 226(4), iyae006. https://doi.org/10.1093/genetics/iyae006

Stankowski, S., Chase, M. A., Fuiten, A. M., Rodrigues, M. F., Ralph, P. L., & Streisfeld, M. A. (2019). Widespread selection and gene flow shape the genomic landscape during a radiation of monkeyflowers. PLOS Biology, 17(7), e3000391. https://doi.org/10.1371/journal.pbio.3000391

Unneberg, P., Larsson, M., Olsson, A., Wallerman, O., Petri, A., Bunikis, I., Vinnere Pettersson, O., Papetti, C., Gislason, A., Glenner, H., Cartes, J. E., Blanco-Bercial, L., Eriksen, E., Meyer, B., & Wallberg, A. (2024). Ecological genomics in the Northern krill uncovers loci for local adaptation across ocean basins. Nature Communications, 15(1), 6297. https://doi.org/10.1038/s41467-024-50239-7

van der Valk, T., Pečnerová, P., Díez-del-Molino, D., Bergström, A., Oppenheimer, J., Hartmann, S., Xenikoudakis, G., Thomas, J. A., Dehasque, M., Sağlıcan, E., Fidan, F. R., Barnes, I., Liu, S., Somel, M., Heintzman, P. D., Nikolskiy, P., Shapiro, B., Skoglund, P., Hofreiter, M., … Dalén, L. (2021). Million-year-old DNA sheds light on the genomic history of mammoths. Nature, 591(7849), 265–269. https://doi.org/10.1038/s41586-021-03224-9

Webster, M. T., Beaurepaire, A., Neumann, P., & Stolle, E. (2023). Population Genomics for Insect Conservation. Annual Review of Animal Biosciences, 11(1), 115–140. https://doi.org/10.1146/annurev-animal-122221-075025

Wetterstrand, KA. DNA Sequencing Costs: Data from the NHGRI Genome Sequencing Program (GSP). www.genome.gov/sequencingcostsdata