...

Snakefile

@@ -44,16 +44,17 @@ rule data:
 
 rule dosage:
     input:
-        "{out}/imputed-genotypes/chromosome-{i}.vcf.gz",
-        "{out}/GTEx_v8_hg38p7_variant_lookup_table.txt.gz"
+        vcf    = "{out}/imputed-genotypes/chromosome-{i}.vcf.gz",
+        lookup = "{out}/GTEx_v8_hg38p7_variant_lookup_table.txt.gz"
     output:
-        "{out}/dosages/chromosome-{i}.dosage.txt.gz"
+        "{out}/dosages/chromosome-{i}-gtex-v8.dosage.txt.gz",
+        "{out}/reports/map-hg38-dosage-to-gtex-variants-chromosome-{i}.html"
     shell:
         """
         set -ex
         mkdir -p {config[output_dir]}/dosages
         # filter for SNPs of defined quality and extract dosage
-        pv {input} | \
+        pv {input.vcf} | \
             bcftools filter -e 'MAF[0]<{config[min_MAF]} | INFO<{config[min_INFO]} | TYPE!="snp" | N_ALT!=1' | \
             bcftools +fill-tags | \
             bcftools query -f \ '%CHROM %ID %POS %REF %ALT %INFO/MAF [%DS ]\n' > \
@@ -61,6 +62,9 @@ rule dosage:
         # compress
         gzip {config[output_dir]}/dosages/chromosome-{wildcards.i}.dosage.txt
         # convert locations to GTEx v8 by hg38 position
+        mkdir -p {wildcards.out}/reports
+        Rscript -e "rmarkdown::render('scripts/map-hg38-dosage-to-gtex-variants.Rmd', knit_root_dir = getwd(), output_dir = '{wildcards.out}/reports', output_file= 'map-hg38-dosage-to-gtex-variants-chromosome-{wildcards.i}', params = list(chromosome = {wildcards.i}))"
+        rm {config[output_dir]}/dosages/chromosome-{wildcards.i}.dosage.txt.gz
         """
 
 # compute dosage for all 23 chromosomes
@@ -68,16 +72,15 @@ rule dosages:
     input:
         rules.data.output,
         expand(
-            "{out}/imputed-genotypes/chromosome-{i}.vcf.gz",
+            "{out}/dosages/chromosome-{i}-gtex-v8.dosage.txt.gz",
             out=config["output_dir"],
             i=range(1,24)
         )
 
 
-
 rule samples_file:
     input:
-        expand("{out}/imputed-genotypes/chromosome-22.vcf", out=config["output_dir"])
+        expand("{out}/imputed-genotypes/chromosome-22.vcf.gz", out=config["output_dir"])
     output:
         expand("{out}/dosages/samples.txt", out=config["output_dir"])
     shell:

config/snakemake/mrc-bsu-cluster/config.yaml

-jobs:               20
+jobs:               25
 use-singularity:    True
 cluster:            "sbatch -A MRC-BSU-SL2-CPU -p skylake --ntasks 1 --cpus-per-task 1 --nodes 1 -t 02:00:00 --job-name '{rule}__{wildcards}' --output '{rule}__{wildcards}.out' --error '{rule}__{wildcards}.err'"
 singularity-args:   "-H $PWD"

scripts/map-hg38-dosage-to-gtex-variants.Rmd

@@ -26,7 +26,9 @@ library(glue, warn.conflicts = FALSE)
 set.seed(42)
 ```
 
-```{r load-data}
+## Chromosome `r params$chromosome`
+
+```{r load-data, message=FALSE}
 dosages_file <- glue(
     "output/dosages/chromosome-{params$chromosome}.dosage.txt.gz"
 )
@@ -35,7 +37,7 @@ out_file <- str_replace(
         dosages_file, 
         "(?<=-)([0-9]{1,2})(?=.dosage.txt)", 
         "\\1-gtex-v8"
-)
+    )
 
 tbl_gtex_lookup <- vroom::vroom(
         "output/GTEx_v8_hg38p7_variant_lookup_table.txt.gz"
@@ -43,7 +45,7 @@ tbl_gtex_lookup <- vroom::vroom(
     filter(
         # to make sure we do not lose corner cases, 
         # also load neighboring chromosomes
-        chr %in% glue("chr{(params$chromosome - 1):(params$chromosome + 1)}"),
+        chr %in% glue("chr{params$chromosome}"),
         nchar(ref) == 1,
         nchar(alt) == 1
     ) %>%
@@ -75,7 +77,7 @@ tbl_dosage <- vroom::vroom(
         col_types = cols(
             .default    = col_double(),
             dummy_      = col_character(),
-            chromosome = col_integer(),
+            chromosome  = col_character(),
             rsid        = col_character(),
             variant_pos = col_integer(),
             ref         = col_character(),
@@ -88,7 +90,7 @@ tbl_dosage <- vroom::vroom(
 ```
 
 
-```{r}
+```{r match-by-position}
 tbl_dosage <- left_join(
         tbl_dosage,
         tbl_gtex_lookup %>% transmute(
@@ -107,37 +109,126 @@ tbl_dosage <- left_join(
     )
 ```
 
-```{r}
-tbl_dosage %>% 
+
+### Unmatched variants
+
+```{r plot-unmatched}
+tbl_unmatched <- tbl_dosage %>% 
     filter(
-        is.na(ref_gtex) | is.na(alt_gtex)
+        is.na(gtex_id)
     ) %>% 
-    {n_mis <<- nrow(.); .} %>% 
-    ggplot() +
+    {n_mis <<- nrow(.); .} 
+
+ggplot(tbl_unmatched) +
     aes(variant_pos) +
     geom_histogram(
         fill  = "black", 
         color = NA,
         bins  = ceiling(n_mis / 33)
     ) + 
+    scale_x_continuous("position") +
+    ggtitle("Histogram of unmatched variants") +
     theme_bw()
-```
 
-```{r}
-tbl_dosage %>% 
-    filter(is.na(ref_gtex) | is.na(alt_gtex)) %>% 
-    ggplot() +
+ggplot(tbl_unmatched) +
     aes(variant_pos, y = 0) +
     geom_point(
         alpha = 0.2, 
         shape = 16, 
         position = position_jitter(width = 0, height = 1)
     ) + 
+    scale_x_continuous("position") +
+    scale_y_continuous("") +
+    ggtitle("Jitter-plot of individual unmatched variants") +
     theme_bw()
 ```
 
+Overall, `r nrow(tbl_unmatched)` variants 
+(`r sprintf("%.1f", 100*nrow(tbl_unmatched)/nrow(tbl_dosage))`%)
+cannot be matched to GTEx v8 variants by position.
+
+```{r match-by-bases}
+tbl_matched <- tbl_dosage %>%
+    filter(
+        !is.na(gtex_id)
+    ) %>%
+    group_by(variant_pos, ref, alt) %>%
+    mutate(
+        match = pmap_lgl(
+            list(ref, alt, ref_gtex, alt_gtex, chromosome, gtex_id),
+            function(ref, alt, ref_gtex, alt_gtex, chromosome, gtex_id) {
+                # match by bases
+                matched <- (ref == ref_gtex) & (alt == alt_gtex)
+                inds    <- which(matched)
+                if (length(inds) > 1) {
+                    matching_chromosome <- chromosome[inds] == str_extract(gtex_id[inds], "^chr[1-9]{1,2}")
+                    if (sum(matching_chromosome) == 0) 
+                        warning("no matching chromosome found")
+                    matched[inds] <- matching_chromosome
+                }
+                return(matched)
+            }
+        )
+    )
+```
+
+```{r mismatch}
+tbl_mismatch <- tbl_matched %>%
+    filter(
+        !any(match)
+    ) %>% 
+    select(
+        chromosome, variant_pos, gtex_id, rsid, ref, alt, ref_gtex, alt_gtex, MAF
+    ) 
 
+pander::pander(tbl_mismatch, caption = "mismatched variants by bases")
+```
 
+Of the position-matched variants `r nrow(tbl_mismatch)`
+have inconsitent bases with GTex v8.
+
+Automatic check that reference base is consistent 
+(no issues with complementing strands etc.): 
+`r assertthat::assert_that(all(tbl_mismatch$ref == tbl_mismatch$ref_gtex))`
+
+```{r check-non-unique}
+tbl_non_unique <- tbl_matched %>%
+    filter(
+        sum(match) > 1
+    ) %>% 
+    select(
+        chromosome, variant_pos, gtex_id, rsid, ref, alt, 
+        ref_gtex, alt_gtex, MAF
+    )
+```
+
+Automatic check that matching for variants is unique: 
+`r assertthat::assert_that(nrow(tbl_non_unique) == 0)`
+
+
+## Matched variants
+
+```{r filter-unique-matches}
+tbl_unique_matched <- tbl_matched %>% 
+    filter(
+        match
+    ) %>% 
+    select(
+        -rsid, -ref_gtex, -alt_gtex
+    ) %>% 
+    ungroup()
+```
+
+A total of 
+`r nrow(tbl_unique_matched)` 
+variants in chromosome 
+`r params$chromosome` 
+could be matched uniquely to a GTEx v8 variant
+(`r sprintf("%.1f", 100*nrow(tbl_unique_matched)/nrow(tbl_dosage))`%).
+
+```{r save}
+write_delim(tbl_unique_matched, out_file)
+```
 
 
 ## Session info