This repository is part of Henry Arenas-Castro's PhD thesis dissertation at the University of Queensland and contains step-by-step instructions to estimate the fertilisation outcome of competitive crosses from PoolSeq data. Maddie E. James assisted with data analysis and Daniel Ortiz-Barrientos and Jan Engelstädter supervised this project.
In absence of fertilisation biases, it is expected that when two sires simultaneously mate or pollinate a dam, each sire fertilises half of the available ovules. Consequently, the estimation of paternity share in the offspring of competitive crosses may be used to test for deviations in the fertilisation outcome, proven that other technical and biological factors are controlled (see below).
Traditional approaches have estimated the paternity share of the offspring by genotyping each individual. However, this is infeasible in large-scale studies. Here, we develop a method to estimate the paternity share in the offspring of competitive crosses using PoolSeq data. Instead of individually genotyping the offspring, we pooled equal amount of tissue from each individual, homogenise them, extracted DNA, prepared RAD-seq libraries, and sequenced a single sample per cross.
To infer the paternity share of the offspring, we (i) identified paternity markers based on the genetic variation of the populations of the parentals and (ii) estimated the allelic frequency of these markers in the offspring.
In the case of competitive crosses between sires from different populations, the genetic markers that are most informative about the paternity of the offpsring are those that both are monomorphic in each population and have different alleles fixed between the two population. These markers should also be monomorphic in the dam population. Depending on whether the dam population shares the same allele with one of the sire populations or not, the expected allelic frequencies in the offspring vary.
All the parental individuals used in the competitive crosses (N=139) were individually genotyped following James et al. (2021) RAD-seq protocol. Then, we jointly called the SNPs for each set of individuals used either as dams or sires per population to identify the sites that were monomorphic. Finally, we identified those sites that had differentially fixed alleles between the sire populations.
We used bwa to index the Senecio lautus reference genome version SPD_CN1K_CtgRN.
bwa index Senecio.contigs.fasta
For each individual, we aligned their reads to the reference genome with the BWA-MEM algorithm. The *_1.fq.gz and *_2.fq.gz files correspond to the forward and reverse read files for each individual. We further used samtools to filter for mapping quality -q 20.
bwa mem -M -t 12 reference_genome.fasta ind1_1.fq.gz ind1_2.fq.gz | samtools view -q 20 -b | samtools sort -@ 12 -T ind1 > ind1_sorted.bam
We also used samtoools to produce a summary of alignment stats per individual, including mean depth and the number and percentage of reads mapped to the reference genome.
samtools depth ind1_sorted.bam | awk '{sum+=$3; sumsq+=$3*$3} END { print sum/NR,"\t", sqrt(sumsq/NR - (sum/NR)**2)}' > ind1_bamdepth.txt
samtools flagstat -O tsv ind1_sorted.bam > ind1_bamstats.txt
To clean the bam files and add read groups to each individual, we used PicardTools, where RGID is the read group ID (i.e. the name of the individual), RGLB is the read group library (i.e. the sequencing lane number), RGPU is the read group platform unit (i.e. run barcode of the lane), and RGSM is the read group sample name (i.e. the name of the individual).
java -jar picard.jar CleanSam INPUT=ind1_sorted.bam OUTPUT=ind1_sorted_cleaned.bam
java -jar picard.jar AddOrReplaceReadGroups INPUT=ind1_sorted_cleaned.bam OUTPUT=ind1_sorted_cleaned_rg.bam SORT_ORDER=coordinate RGID=ind1 RGLB=lib1 RGPL=ILLUMINA RGPU=lane1_parentals RGSM=ind1 CREATE_INDEX=True
We used FreeBayes to jointly call SNPs for each set of dams and sires per population. Joint calling assumes all samples are genetically similar and uses information across all samples to call a SNP. SNPs from these cohorts are then combined into one VCF file. The code below shows an example where ind1, ind2, and ind3 conform the set of dams from pop1.
freebayes -f Senecio.contigs.fasta ind1_sorted_cleaned_rg.bam ind2_sorted_cleaned_rg.bam ind3_sorted_cleaned_rg.bam --use-best-n-alleles 4 --report-monomorphic --genotype-qualities > pop1_dams.vcf
For each cross, or pool, we used bcftools view to subsample the individuals that effectively served as sires and dams. The -s flag specifies the name of the individuals to sample from the population's VCF (i.e. only ind1 and ind3 from the step above).
bcftools view -s ind1,ind3 pop1_dams.vcf > pop1_dams_ind.vcf
Before starting to filter the SNPs, we used bgzip and tabix to compress and index each VCF.
bgzip -c pop1_dams_ind.vcf > pop1_dams_ind.vcf.gz
tabix -p vcf pop1_dams_ind.vcf.gz
To identify and retain the sites that vary among the three parental populations, we used bcftools merge and vcftools to merge the VCFs and keep the sites genotyped in at least 50% of individuals with a minimum quality score of 30.
bcftools merge pop1_dams_ind.vcf.gz pop2_sires_ind.vcf.gz pop3_sires_ind.vcf.gz -o Pool1.vcf
vcftools --gzvcf Pool1.vcf --max-missing 0.5 --mac 1 --minQ 30 --recode --recode-INFO-all --stdout | gzip -c > Pool1_Q30mac1.vcf.gz
We applied a minimum depth per sample of 3 reads (genotypes with less than 3 reads were recoded as missing data).
vcftools --gzvcf Pool1_Q30mac1.vcf.gz --minDP 3 --recode --recode-INFO-all --stdout > Pool1_Q30mac1dp3.vcf
Sites with high coverage could be multi-copy regions of the genome, so we removed these potential paralogues. We first calculated the mean depth per site after pooling the VCFs of all the individuals used in this study and plot their distribution in R (min=1.8, max=2416.3). Then, we visually examined the distribution of mean read depth across all sites and selected a threshold value of 250.
vcftools --vcf Pool1_Q30mac1dp3.vcf --site-mean-depth --out mean_depth
vcftools --vcf Pool1_Q30mac1dp3.vcf --recode --recode-INFO-all --out Pool1_Q30mac1dp3MaxDP250 --max-meanDP 250
We then filtered for a minimum mean depth of 10.
vcftools --vcf Pool1_Q30mac1dp3MaxDP250.recode.vcf --min-meanDP 10 --recode --recode-INFO-all --out Pool1_Q30mac1dp3MaxDP250MinDP10
We additionally filtered for an overall missing data, removing sites if they had greater than 20% missing data.
vcftools --vcf Pool1_Q30mac1dp3MaxDP250MinDP10.recode.vcf --recode --recode-INFO-all --max-missing 0.8 --out Pool1_Q30mac1dp3MaxDP250MinDP10md80
Finally, we used plink to split the merged VCF into the original dam and sire populations. To specify the name of the individuals that composed each population, we prepared a tab-delimited three-column popmap file per population. The first two columns had the name of the individual and the third column the number 1. Each file contained one row per individual.
plink --vcf Pool1_Q30mac1dp3MaxDP250MinDP10mpp20md80.recode.vcf --keep pop1_dams_popmap.txt --export vcf --allow-extra-chr --out pop1_dams_filtered
We calculated the allelic frequencies per loci in each parental population per pool using plink2.
plink2 --vcf pop1_dams_filtered.vcf --freq --allow-extra-chr --out Pool1_dam --set-all-var-ids @:#
We then used R to identify the paternity markers across the 226 pools of this study.
# Emtpy table to summarise results
results <- matrix(NA, nrow=226, ncol=2)
# Repeat this operation across all pools
for (i in 1:226) {
# DAM
# Open file with the allelic frequencies per site
dam <- read.csv(paste("Pool", i, "_dam.afreq", sep=""), sep="", header=T) # Any lenght white space delimiter
# Remove sites with missing data
dam <- dam[dam[,6]==max(dam[,6]),]
# Remove sites with indels
dam <- dam[dam[,3]=="A" | dam[,3]=="G" | dam[,3]=="C" | dam[,3]=="T",]
dam <- dam[dam[,4]=="A" | dam[,4]=="G" | dam[,4]=="C" | dam[,4]=="T",]
# Remove polymorphic sites (MAC=0)
dam <- dam[dam[,5]==0,]
# Keep only location (col 2) and major allele (col 3)
dam <- dam[,c(2,3)]
# SIRE 1
sire1 <- read.csv(paste("Pool", i, "_sire1.afreq", sep=""), sep="", header=T) # Any lenght white space delimiter
# Remove sites with missing data
sire1 <- sire1[sire1[,6]==max(sire1[,6]),]
# Remove sites with indels
sire1 <- sire1[sire1[,3]=="A" | sire1[,3]=="G" | sire1[,3]=="C" | sire1[,3]=="T",]
sire1 <- sire1[sire1[,4]=="A" | sire1[,4]=="G" | sire1[,4]=="C" | sire1[,4]=="T",]
# Remove polymorphic sites (MAC=0)
sire1 <- sire1[sire1[,5]==0,]
# Keep only location (col 2) and major allele (col 3)
sire1 <- sire1[,c(2,3)]
# SIRE 2
sire2 <- read.csv(paste("Pool", i, "_sire2.afreq", sep=""), sep="", header=T) # Any lenght white space delimiter
# Remove sites with missing data
sire2 <- sire2[sire2[,6]==max(sire2[,6]),]
# Remove sites with indels
sire2 <- sire2[sire2[,3]=="A" | sire2[,3]=="G" | sire2[,3]=="C" | sire2[,3]=="T",]
sire2 <- sire2[sire2[,4]=="A" | sire2[,4]=="G" | sire2[,4]=="C" | sire2[,4]=="T",]
# Remove polymorphic sites (MAC=0)
sire2 <- sire2[sire2[,5]==0,]
# Keep only location (col 2) and major allele (col 3)
sire2 <- sire2[,c(2,3)]
# Monomorphic sites across sire populations
mono <- intersect(sire1[,1], sire2[,1])
sire1 <- sire1[which(sire1[,1] %in% mono),]
sire2 <- sire2[which(sire2[,1] %in% mono),]
# Merge
sire12 <- merge(sire1, sire2, by=1)
sire12 <- droplevels(sire12)
# Differentially fixed alleles
diff <- sire12[which(sire12[,2] != sire12[,3]),]
# Common monomorphic between the dam and the sires
dam <- dam[which(dam[,1] %in% diff[,1]),]
trio <- merge(diff, dam, by=1)
colnames(trio) <- c("POSITION", "SIRE1", "SIRE2", "DAM")
# Output results
write.table(trio, paste("Pool", i, "_markers.txt", sep=""),
quote=F, row.names=F, sep="\t", col.names=T)
results[i,] <- c(paste("Pool", i, sep=""), nrow(trio))
}
# Save results
colnames(results) <- c("POOL", "MARKERS")
write.table(results, "PaternityMarkers.txt", quote=F, row.names=F, sep="\t", col.names=T)
The number of paternity markers per pool ranged from 1 to 410, with an average value of 101.
Similar to the parental individuals, all the pools (N=226) were individually genotyped following James et al. (2021) RAD-seq protocol. We also mapped the reads to the reference genome using BWA-MEM and removed ambiguosly mapped reads with samtools.
Since each pool contains DNA from multiple individuals, we used PoPoolation2 to call the SNPs. This is a reliable program to calculate allelic frequencies from pooled DNA samples.
We jointly called the SNPs by pool pairs. For this, we firt created a pileup file for each pair using samtools and then synchronised it with mpileup2sync.
samtools mpileup -B pool1_sorted.bam pool2_sorted.bam > pool1-2.mpileup
java -ea -Xmx7g -jar mpileup2sync.jar --input pool1-2.mpileup --output pool1-2.sync --fastq-type sanger --min-qual 20 --threads 12
This file contains the allele frequencies of each pool across all the mapped positions and serves as input for PoPoolation2.
perl snp-frequency-diff.pl --input pool1-2.sync --output-prefix pool1-2 --min-count 6 --min-coverage 25 --max-coverage 250
PoPoolation2 output two files: *_rc and *_pwc. The former one contains the count of the major and minor allele in a concise format:
##chr pos rc allele_count allele_states deletion_sum snp_type major_alleles(maa) minor_alleles(mia) maa_1 maa_2 mia_1 mia_2
SPDCN1KCT_24 9367 N 2 C/G 0 pop CC GG 21/28 24/28 7/28 4/28
SPDCN1KCT_24 9404 N 2 A/C 0 pop AC NA 28/28 19/28 0/28 9/28
SPDCN1KCT_24 9459 N 2 G/T 0 pop TG NN 20/20 21/21 0/20 0/21
We used R to match the paternity markers with the SNPs effectively called in the pools and extract their allelic frequencies.
# Plot histograms
par(mfrow=c(3,6), mar=c(3,5,7,3)) # CHANGE ACCORDING THE NUMBER OF POOLS
# Emtpy table to summarise results
results <- matrix(NA, nrow=226, ncol=8)
# Doing this in all pairs, from pool1-2 to pool225-226
for (s in seq(1, 225, 2)) {
# Position of the pool in the PoPoolation2 output
order <- c(paste("Pool", s, sep=""), paste("Pool", s+1, sep=""))
for (p in 1:length(order)) {
# List of paternity markers per pool
marker <- read.csv(paste("~/Dropbox/_PhD_OB/Data/9_PaternityAssignment/afreq/", order[p], "_markers.txt", sep=""), sep="\t", header=T)
# PoPoolation2 output file (*_rc)
pool <- read.csv(paste("pool",s,"-",s+1,"_rc", sep=""), sep="\t", header=T)
# Create a new column in the pool matrix with the position of the SNP as it is in the marker matrix
pool$POSITION <- paste(pool[,1], pool[,2], sep=":")
# Paternity markers called as SNPs in the pools
pool <- pool[which(pool$POSITION %in% marker$POSITION),]
# Keep only these columuns: POSITION, ALLELE COUNT, ALLELE STATES, MAJOR ALLELE, MINOR ALLELE, MAJOR ALLELE COUNT
pool <- pool[,c(ncol(pool), 4, 5, 8, 9, 9+p)]
# Get the right major and minor allele for the pool instead of a block for all the pools in the group
pool$major_alleles.maa. <- as.character(pool$major_alleles.maa.) # Major allele
pool$minor_alleles.mia. <- as.character(pool$minor_alleles.mia.) # Minor allele
for (i in 1:nrow(pool)) {
pool$major_alleles.maa.[i] <- strsplit(pool$major_alleles.maa.[i], "")[[1]][p]
pool$minor_alleles.mia.[i] <- strsplit(pool$minor_alleles.mia.[i], "")[[1]][p]
}
# Merge the paternity marker and pool information
pool <- merge(pool, marker, by=1)
# Calculate the allelic frequencies (Dam allele as reference)
pool[,6] <- as.character(pool[,6]) # Allelic count
pool[,9] <- as.character(pool[,9]) # Dam allele
pool$freq <- NA # Create a new column
for (j in 1:nrow(pool)) {
# Check that the major allele corresponds to the dam's allele
if (pool[j,4] == pool[j,9]) {
afreq <- strsplit(pool[j,6], "/")
pool$freq[j] <- as.numeric(afreq[[1]][1])/as.numeric(afreq[[1]][2])
}
# Ohterwise output the minor allele frequency
else {
pool$freq[j] <- 1-(as.numeric(afreq[[1]][1])/as.numeric(afreq[[1]][2]))
}
}
# Note whether the dam share its allele with sire1 or sire2
pool$DAM_ALLELE <- NA
pool[which(pool[,9] == pool[,7]), 11] <- "sire1"
pool[which(pool[,9] == pool[,8]), 11] <- "sire2"
# Save the results per pool
write.table(pool, paste(order[p], "_FreqMarkers.txt", sep=""), quote=F, row.names=F, sep="\t", col.names=T)
# Allele frequency distribution per pool
boxplot(pool[pool$DAM_ALLELE=="sire1",10], pool[pool$DAM_ALLELE=="sire2",10], ylim=c(0,1), cex.main=2, axes=F,
main=paste(order[p], "\nN=", length(pool[pool$DAM_ALLELE=="sire1",10]),"+",length(pool[pool$DAM_ALLELE=="sire2",10]), sep=""))
axis(2,cex.axis=2)
abline(h=0.75, lty="longdash", col="grey50", lwd=1.5)
# Summarise the results across all pools
# Col 1: Pool name
# Col 2: Total number markers
# Col 3: Number of dam alleles similar to sire1
# Col 4: Mean allelic freq. of dam alleles similar to sire1
# Col 5: SD of mean allelic freq. of dam alleles similar to sire1
# Col 6: Number of dam alleles similar to sire2
# Col 7: Mean allelic freq. of dam alleles similar to sire2
# Col 8: SD of mean allelic freq. of dam alleles similar to sire2
results[s+p-1,] <- c(order[p], nrow(pool),
length(pool[pool$DAM_ALLELE=="sire1",10]), mean(pool[pool$DAM_ALLELE=="sire1",10]), sd(pool[pool$DAM_ALLELE=="sire1",10]),
length(pool[pool$DAM_ALLELE=="sire2",10]), mean(pool[pool$DAM_ALLELE=="sire2",10]), sd(pool[pool$DAM_ALLELE=="sire2",10]))
}
}
# Save summary of results
colnames(results) <- c("POOL", "MARKERS", "N_SIRE1", "MEAN_SIRE1", "SD_SIRE1", "N_SIRE2", "MEAN_SIRE2", "SD_SIRE2")
write.table(results, "PaternityResults.txt", quote=F, row.names=F, sep="\t", col.names=T)