Show code
library(rtracklayer)
library(GenomicRanges)
library(GenomicFeatures)
library(ggplot2)
library(AnnotationDbi)
library(dplyr)
library(tidyr)
library(ggsci)
library(ggrastr)
library(patchwork)
library(rstatix)
library(ggpubr)SF3B1 splicing analysis
Exon regulation
1 Analysis description
This report holds all analysis and plots used to create main figure 5 D-F with and all related supplementary figures.
1.1 Load libraries
Show code
source("../styles.R")
source("../helper.R")2 Alternative AGs at 3’ splice sites
Show code
ss3PacBioEnd_1 = readRDS("../02_splicingRegulation/data/ss3PacBioEnd_1.rds")
ss3PacBioEnd_2 = readRDS("../02_splicingRegulation/data/ss3PacBioEnd_2.rds")
ss3PacBioEnd = readRDS("../02_splicingRegulation/data/ss3PacBioEnd.rds")
grAgPacBio = readRDS("../02_splicingRegulation/data/grAgPacBio.rds")
maxEntRes = read.csv("../02_splicingRegulation/data/seqsAGsPacBio_maxEnt.fasta", header = FALSE, sep = "\t")
maxEntResScore = sapply(strsplit(maxEntRes$V2,":"), `[`, 2)
maxEntResScore = maxEntResScore[!is.na(maxEntResScore)]
maxEntResScore = as.numeric(maxEntResScore)
maxEntResExonID = sapply(strsplit(maxEntRes$V1,"_"), `[`, 1)
maxEntResExonID = maxEntResExonID[grepl(">", maxEntResExonID)]
maxEntResExonID = gsub(x = maxEntResExonID, pattern = ">", replacement = "")
maxEntAgID = maxEntRes$V1
maxEntAgID = maxEntAgID[grepl(">", maxEntAgID)]
maxEntAgID = gsub(x = maxEntAgID, pattern = ">", replacement = "")
maxEntRes = data.frame(exonID = maxEntResExonID, agID = maxEntAgID, score = maxEntResScore)
grAgPacBio$maxEntScore = maxEntRes$score[match(grAgPacBio$combID, maxEntRes$agID)]
### ----------------------------------------------------------------------------
### Add PacBio splicing results to PacBio AGs
### ----------------------------------------------------------------------------
###
idx = match(grAgPacBio$exonID, ss3PacBioEnd$exonID)
grAgPacBio$regulation = ss3PacBioEnd$regulation[idx]
grAgPacBio$sig = ss3PacBioEnd$sig[idx]
grAgPacBio$sigAlt = ifelse(grAgPacBio$isAlternative == "Yes" & grAgPacBio$sig == TRUE, "Yes", "No")For this analysis we used all significant 3’ alternative splicing events detected in the IsoSeq analysis (N = 7,951). The ranges of these events were reduced to unique 3’ splice site positions (N = 7,167). For overlapping events, the one with the lowest adjusted P value is preferred. Resulting 3’SS that do not have a classical AG splice site were removed as well, leaving a final set of 6,992 splice sites.
The final set is split by whether the events elongates or shortens the exons. This regulation type is defined based on the deltaPSI value from the IsoSeq analysis. A positive value indicates that introns become shorter and a negative value indicates that introns become longer (N = 3,409, 3,583, respectively).
2.1 Splice site strength by MaxEntScan
The splice site strength analysis was performed with all putative AGs for each of the detected 3’splice sites. That is all AGs present in a 150nt window (100nt into the intron and 50nt into the exon) from the detected splice site.
MaxEntScan requires a specific sequence of 23nt around each site has to be used. This sequence consists of exactly 20nt into the intron plus 3nt of the exon.
Show code
### ----------------------------------------------------------------------------
### Make MaxEnt splicing map
### ----------------------------------------------------------------------------
###
dfMain = mcols(grAgPacBio) %>%
as.data.frame() %>%
mutate(agDist = ifelse(agID2 >= 0, -(99-agPosition), agPosition)) %>%
mutate(color = ifelse(sigAlt == "Yes", "Evt sig + Alt AG",
ifelse(sigAlt == "No" & isAlternative == "Yes", "Evt not sig + Alt AG", "Evt not sig + Other AG"))) %>%
mutate(color = factor(color, levels = c("Evt not sig + Other AG", "Evt not sig + Alt AG", "Evt sig + Other AG", "Evt sig + Alt AG"))) %>%
arrange(color) %>%
filter(regulation == "shorter_introns")
dfCanonicalSites = dfMain %>% filter(agID2 == 0)
p0 = ggplot() +
geom_point_rast(data = dfCanonicalSites, aes(x = agDist, y = maxEntScore),
color = "#F2D388",size = 4, alpha = 0.5) +
geom_point(data = dfMain, aes(x = agDist, y = maxEntScore, color = color),
shape = 1, size = 1) +
theme_pub() +
annotate(geom = "rect", xmin = -100, xmax = 50, ymin = -59, ymax = -61) +
annotate(geom = "rect", xmin = 0, xmax = 50, ymin = -55, ymax = -65) +
annotate(geom = "rect", xmin = 0-21, xmax = 0-12, ymin = -Inf, ymax = Inf, color = "black", fill = NA, alpha = 0.5) +
theme(legend.position = "bottom") +
scale_color_nice_map_b() +
labs(
x = "Position of putative and used alternative 3' splice sites relative to the canonical [nt]",
y = "Splice site strength (MaxEntScore)",
color = "Legend") +
guides(colour = guide_legend(override.aes = list(size=5, shape = 15)))
### ----------------------------------------------------------------------------
### Bin AG position and add histograms
### ----------------------------------------------------------------------------
###
bins = seq(from = -99, to = 53, by = 3)
matchedBinInterval = findInterval(dfMain$agDist, bins, all.inside = TRUE, rightmost.closed = TRUE)
dfBin1 = dfMain %>%
mutate(bin = bins[matchedBinInterval]) %>%
group_by(bin) %>%
tally()
p1 = ggplot(dfBin1, aes(x = bin, y = (n))) +
geom_col(position = "dodge", fill = "#808080") +
theme_pub() +
theme(legend.position = "none") +
theme(axis.title.x=element_blank()) +
labs(y = "#N")
dfBin2 = dfMain %>%
mutate(bin = bins[matchedBinInterval]) %>%
group_by(bin, color) %>%
tally() %>%
mutate(color = as.factor(color))
p2 = ggplot(dfBin2, aes(x = bin, y = (n), fill = color)) +
geom_col(position = "fill") +
scale_fill_nice_map_b() +
theme_pub() +
theme(legend.position = "none") +
theme(axis.title.x=element_blank()) +
labs(y = "Proportion")
### ----------------------------------------------------------------------------
### Make combine plot
### ----------------------------------------------------------------------------
###
(p1 / p2 / p0) +
plot_layout(heights = c(2,2,6)) +
plot_annotation(title = "3'splice sites from PacBio defined exons",
subtitle = paste0("Total of ", myFormat(nrow(dfMain)),
" AGs from ", myFormat(length(unique(dfMain$exonID))),
" events, with ", myFormat(nrow(subset(dfMain, isAlternative == "Yes"))),
" used alternative AGs from ", myFormat(length(unique(dfMain$exonID[dfMain$isAlternative == "Yes"]))),
" events.\nAnd ", myFormat(nrow(subset(dfMain, sigAlt == "Yes"))),
" used alternative AGs that are significant from ", myFormat(length(unique(dfMain$exonID[dfMain$sigAlt == "Yes"]))),
" events."
))
Show code
ggsave("maxEntPlot_v1.pdf", width = 8, height = 8, device = "pdf", path = "./plots/")Show code
### ----------------------------------------------------------------------------
### Make MaxEnt splicing map
### ----------------------------------------------------------------------------
###
dfMain = mcols(grAgPacBio) %>%
as.data.frame() %>%
mutate(agDist = ifelse(agID2 >= 0, -(99-agPosition), agPosition)) %>%
mutate(color = ifelse(sig == TRUE & isAlternative == "Yes", "Evt sig + Alt AG",
ifelse(sig == FALSE & isAlternative == "Yes", "Evt not sig + Alt AG",
ifelse(sig == TRUE & isAlternative == "No", "Evt sig + Other AG", "Evt not sig + Other AG")))) %>%
mutate(color = factor(color, levels = c("Evt not sig + Other AG", "Evt sig + Other AG", "Evt not sig + Alt AG", "Evt sig + Alt AG"))) %>%
arrange(color) %>%
filter(regulation == "shorter_introns")
dfCanonicalSites = dfMain %>% filter(agID2 == 0)
p0 = ggplot() +
geom_point_rast(data = dfCanonicalSites, aes(x = agDist, y = maxEntScore),
color = "#F2D388",size = 4, alpha = 0.5) +
geom_point(data = dfMain, aes(x = agDist, y = maxEntScore, color = color),
shape = 1, size = 1) +
theme_pub() +
annotate(geom = "rect", xmin = -100, xmax = 50, ymin = -59, ymax = -61) +
annotate(geom = "rect", xmin = 0, xmax = 50, ymin = -55, ymax = -65) +
annotate(geom = "rect", xmin = 0-21, xmax = 0-12, ymin = -Inf, ymax = Inf, color = "black", fill = NA, alpha = 0.5) +
theme(legend.position = "bottom") +
scale_color_nice_full_b() +
labs(
x = "Position of putative and used alternative 3' splice sites relative to the canonical [nt]",
y = "Splice site strength (MaxEntScore)",
color = "Legend") +
guides(fill=guide_legend(ncol=2), color = guide_legend(ncol=2)) +
guides(colour = guide_legend(override.aes = list(size=5, shape = 15)))
### ----------------------------------------------------------------------------
### Bin AG position and add histograms
### ----------------------------------------------------------------------------
###
bins = seq(from = -99, to = 53, by = 3)
matchedBinInterval = findInterval(dfMain$agDist, bins, all.inside = TRUE, rightmost.closed = TRUE)
dfBin1 = dfMain %>%
mutate(bin = bins[matchedBinInterval]) %>%
group_by(bin) %>%
tally()
p1 = ggplot(dfBin1, aes(x = bin, y = (n))) +
geom_col(position = "dodge", fill = "#808080") +
theme_pub() +
theme(legend.position = "none") +
theme(axis.title.x=element_blank()) +
labs(y = "#N")
dfBin2 = dfMain %>%
mutate(bin = bins[matchedBinInterval]) %>%
group_by(bin, color) %>%
tally() %>%
mutate(color = as.factor(color))
p2 = ggplot(dfBin2, aes(x = bin, y = (n), fill = color)) +
geom_col(position = "fill") +
scale_fill_nice_full_b() +
theme_pub() +
theme(legend.position = "none") +
theme(axis.title.x=element_blank()) +
labs(y = "Proportion")
### ----------------------------------------------------------------------------
### Make combine plot
### ----------------------------------------------------------------------------
###
(p1 / p2 / p0) +
plot_layout(heights = c(2,2,6)) +
plot_annotation(title = "3'splice sites from PacBio defined exons",
subtitle = paste0("Total of ", myFormat(nrow(dfMain)),
" AGs from ", myFormat(length(unique(dfMain$exonID))),
" events.\nWith ", as.character(myFormat(table(dfMain$color)[1])), " ", as.character(names(table(dfMain$color)[1])),
" AGs and, ", as.character(myFormat(table(dfMain$color)[2])), " ", as.character(names(table(dfMain$color)[2])),
" AGs and,\n", as.character(myFormat(table(dfMain$color)[3])), " ", as.character(names(table(dfMain$color)[3])),
" AGs and, ", as.character(myFormat(table(dfMain$color)[4])), " ", as.character(names(table(dfMain$color)[4]))
)
)
Show code
ggsave("maxEntPlot_v2.pdf", width = 8, height = 8, device = "pdf", path = "./plots/")3 Exon AG features (by AG)
Here AG features, like splice site strength and distances, are grouped based on the relative position of the alternative AG. This means from every splicing event, always the first, second, third, etc. AG fall into the same bin.
Note that the first position (AG-1) is probably influenced by the NAGNAG sites or similar splice sites. To avoid a biased distribution, I removed such sites, by filtering for distances of 3,4,5 and 6 nt. This means that all AGs on these position were removed.
3.1 Distances to canonical AG
Here we calculate the distance in nt from the canonical splice site up to the 5th putative alternative AG within the first 50nt to the canonical splice site.
Show code
dfAll = mcols(grAgPacBio) %>%
as.data.frame() %>%
filter(regulation == "shorter_introns") %>%
mutate(agDist = abs(agPosition - 99)) %>%
filter(agID2 > 0 & agID2 <= 5) %>%
mutate(agID2 = factor(agID2)) %>%
select(agPosition, agID2, agDist, isAlternative, maxEntScore, sig, sigAlt) %>%
filter(agDist <= 50)
dfFilter = dfAll %>% filter(!agDist %in% c(3,4,5,6))
dfUsed = dfFilter %>% filter(isAlternative == "Yes") %>%
mutate(agID2 = "Alt")
dfNotUsed = dfFilter %>% filter(isAlternative == "No") %>%
mutate(agID2 = "Not Used")
dfPlot = rbind.data.frame(dfFilter, dfUsed, dfNotUsed)
nC = dfPlot %>%
group_by(sig, agID2) %>%
summarise(count = n()) %>%
mutate(count = myFormat(count))
p = ggplot() +
geom_boxplot(data = dfPlot, aes(x = agID2, y = agDist, fill = sig, color = sig), outlier.size = 0.5) +
theme_pub() +
theme(legend.position = "top") +
scale_fill_nice_gr_b() +
scale_color_nice_gr_d() +
labs(
title = "Distance between canonical and alternative AGs",
subtitle = paste0("NAGNAG sites removed (all=", myFormat(nrow(dfAll)), ", removed=", myFormat(nrow(dfFilter)), "),\nwith P value by T-test and BH."),
x = "Putative AG",
y = "Distance to canonical AG",
color = "Significant event",
fill = "Significant event"
) +
geom_vline(xintercept = 5.5, linetype = "dashed") +
geom_text(data = nC, aes(x = agID2, y = 0, label = count, group = sig), position = position_dodge(width = 0.8), angle = 45, size = 2.5)
statTest = dfPlot %>%
group_by(agID2) %>%
t_test(agDist ~ sig) %>%
adjust_pvalue(method = "BH") %>%
add_significance("p.adj") %>%
add_xy_position(x = "agID2", dodge = 0.8) %>%
mutate(y.position = 55.1)
p + stat_pvalue_manual(data = statTest, label = "p", tip.length = 0.02, size = 3) 
Show code
ggsave("features_distance_bar.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
df = dfPlot %>%
group_by(sig, isAlternative, agID2) %>%
tally() %>%
group_by(sig, agID2) %>%
summarize(sig, isAlternative, agID2, n, grpSum = sum(n)) %>%
arrange(agID2) %>%
mutate(frac = round((n / grpSum) *100, digits = 1)) %>%
filter(isAlternative == "Yes")
ggplot(df, aes(x = agID2, y = frac, fill = sig)) +
geom_col(position = position_dodge(preserve = 'single')) +
theme_pub() +
theme(legend.position = "top") +
scale_fill_nice_gr_b() +
scale_color_nice_gr_d() +
scale_y_continuous(labels = scales::percent_format(scale = 1)) +
labs(
# title = "Proportion of alternative AGs at given position",
x = "Alternative AG",
y = "Proportion of alternative AGs",
color = "Significant event",
fill = "Significant event"
) +
geom_vline(xintercept = 5.5, linetype = "dashed") 
Show code
ggsave("features_distance_box.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")3.2 Splice site strength by AG
Here we calculate the splice site strength with MaxEnt for the first 5 putative alternative AG within the first 50nt to the canonical splice site.
Show code
dfAll = mcols(grAgPacBio) %>%
as.data.frame() %>%
filter(regulation == "shorter_introns") %>%
mutate(agDist = abs(agPosition - 99)) %>%
filter(agID2 >= 0 & agID2 <= 5) %>%
mutate(agID2 = factor(agID2)) %>%
select(agPosition, agID2, agDist, isAlternative, maxEntScore, sig, sigAlt) %>%
filter(agDist <= 50)
dfFilter = dfAll %>% filter(!agDist %in% c(3,4,5,6))
dfUsed = dfFilter %>% filter(isAlternative == "Yes") %>%
mutate(agID2 = "Alt")
dfNotUsed = dfFilter %>% filter(isAlternative == "No") %>%
mutate(agID2 = "Not Used")
dfPlot = rbind.data.frame(dfFilter, dfUsed, dfNotUsed)
nC = dfPlot %>%
group_by(sig, agID2) %>%
summarise(count = n()) %>%
mutate(count = myFormat(count))
p = ggplot() +
geom_boxplot(data = dfPlot, aes(x = agID2, y = maxEntScore, fill = sig, color = sig), outlier.size = 0.5) +
theme_pub() +
theme(legend.position = "top") +
scale_fill_nice_gr_b() +
scale_color_nice_gr_d() +
labs(
title = "Strength of putative AGs",
subtitle = paste0("NAGNAG sites removed (all=", myFormat(nrow(dfAll)), ", removed=", myFormat(nrow(dfFilter)), "),\nwith P value by T-test and BH."),
x = "Putative AG",
y = "Splice site strength (maxEntScore)",
color = "Significant event",
fill = "Significant event"
) +
geom_vline(xintercept = 5.5, linetype = "dashed") +
geom_text(data = nC, aes(x = agID2, y = 0, label = count, group = sig), position = position_dodge(width = 0.8), angle = 45, size = 2.5)
statTest = dfPlot %>%
group_by(agID2) %>%
t_test(maxEntScore ~ sig) %>%
adjust_pvalue(method = "BH") %>%
add_significance("p.adj") %>%
add_xy_position(x = "agID2", dodge = 0.8) %>%
mutate(y.position = 19.1)
p + stat_pvalue_manual(data = statTest, label = "p", tip.length = 0.02, size = 2.5) 
Show code
ggsave("features_maxEnt_box_v1.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
p = ggplot() +
geom_boxplot(data = dfFilter, aes(x = agID2, y = maxEntScore, fill = sig, color = sig), outlier.size = 0.5) +
scale_fill_nice_gg_b() +
scale_color_nice_gg_d() +
ggnewscale::new_scale_color() +
ggnewscale::new_scale_fill() +
geom_jitter(data = subset(dfFilter, isAlternative == "Yes"),
aes(x = agID2, y = maxEntScore, color = sig),
size = 0.2, position = position_jitterdodge()) +
scale_color_nice_br_b() +
theme_pub() +
theme(legend.position = "top") +
guides(fill=guide_legend(nrow=2,byrow=TRUE)) +
labs(
title = "Strength of putative AGs",
subtitle = paste0("NAGNAG sites removed (all=", myFormat(nrow(dfAll)), ", removed=", myFormat(nrow(dfFilter)), ")"),
x = "Putative AG",
y = "Splice site strength (maxEntScore)",
color = "Type",
fill = "Type"
)
statTest = dfFilter %>%
group_by(agID2) %>%
t_test(maxEntScore ~ sig) %>%
adjust_pvalue(method = "BH") %>%
add_significance("p.adj") %>%
add_xy_position(x = "agID2", dodge = 0.8) %>%
mutate(y.position = 19.1)
p + stat_pvalue_manual(data = statTest, label = "p", tip.length = 0.02, size = 2.5) 
Show code
ggsave("features_maxEnt_box_v2.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
dfAll = mcols(grAgPacBio) %>%
as.data.frame() %>%
filter(regulation == "shorter_introns") %>%
mutate(agDist = abs(agPosition - 99)) %>%
filter(agID2 >= 0 & agID2 <= 5) %>%
mutate(agID2 = factor(agID2)) %>%
select(agPosition, agID2, agDist, isAlternative, maxEntScore, sig, sigAlt) %>%
filter(agDist <= 50)
dfFilter = dfAll %>%
filter(!agDist %in% c(3,4,5,6)) %>%
mutate(sigNew = paste0("Alt_",isAlternative, "_sig_", sig))
nC = dfFilter %>%
group_by(sigNew, agID2) %>%
summarise(mScore = mean(maxEntScore), count = n()) %>%
mutate(count = myFormat(count))
ggplot() +
geom_boxplot(data = dfFilter, aes(x = agID2, y = maxEntScore, fill = sigNew, color = sigNew), outlier.size = 0.5) +
scale_fill_nice_full_b() +
scale_color_nice_full_d() +
geom_vline(xintercept = 1.5, linetype = "dashed") +
geom_text(data = nC, aes(x = agID2, y = mScore, label = count, group = sigNew), position = position_dodge(width = 0.8), angle = 45, size = 3) +
theme_pub() +
theme(legend.position = "top") +
guides(fill=guide_legend(nrow=2,byrow=TRUE)) +
labs(
title = "Strength of putative AGs",
subtitle = paste0("NAGNAG sites removed (all=", myFormat(nrow(dfAll)), ", removed=", myFormat(nrow(dfFilter)), ")"),
x = "Putative AG",
y = "Splice site strength (maxEntScore)",
color = "Type",
fill = "Type"
) 
Show code
ggsave("features_maxEnt_box_v3.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
dfFilter = dfFilter %>% mutate(type = ifelse(agID2 == 0, "Canonical", "Putative alternative"))
nC = dfFilter %>%
group_by(sigNew, type) %>%
summarise(mScore = mean(maxEntScore), count = n()) %>%
mutate(count = myFormat(count))
ggplot() +
geom_boxplot(data = dfFilter, aes(x = type, y = maxEntScore, fill = sigNew, color = sigNew), outlier.size = 0.5) +
scale_fill_nice_full_b() +
scale_color_nice_full_d() +
geom_vline(xintercept = 1.5, linetype = "dashed") +
geom_text(data = nC, aes(x = type, y = mScore, label = count, group = sigNew), position = position_dodge(width = 0.8), angle = 45, size = 3) +
theme_pub() +
theme(legend.position = "top") +
guides(fill=guide_legend(nrow=2,byrow=TRUE)) +
labs(
title = "Strength of putative AGs",
subtitle = paste0("NAGNAG sites removed (all=", myFormat(nrow(dfAll)), ", removed=", myFormat(nrow(dfFilter)), ")"),
x = "Putative AG",
y = "Splice site strength (maxEntScore)",
color = "Type",
fill = "Type"
) 
Show code
ggsave("features_maxEnt_box_v4.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
df = dfPlot %>%
group_by(sig, isAlternative, agID2) %>%
tally() %>%
group_by(sig, agID2) %>%
summarize(sig, isAlternative, agID2, n, grpSum = sum(n)) %>%
arrange(agID2) %>%
mutate(frac = round((n / grpSum) *100, digits = 1)) %>%
filter(isAlternative == "Yes")
ggplot(df, aes(x = agID2, y = frac, fill = sig)) +
geom_col(position = position_dodge(preserve = 'single')) +
theme_pub() +
theme(legend.position = "top") +
scale_fill_nice_gr_b() +
scale_color_nice_gr_d() +
scale_y_continuous(labels = scales::percent_format(scale = 1)) +
labs(
title = "Proportion of alternative AGs",
x = "Alternative AG",
y = "Proportion of alternative AGs",
color = "Significant event",
fill = "Significant event"
) +
geom_vline(xintercept = 5.5, linetype = "dashed") 
Show code
ggsave("features_maxEnt_bar_v1.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
df = dfFilter %>% filter(agID2 != 0) %>%
group_by(agID2, sigNew, type) %>%
summarise(n = n()) %>% filter(type == "Putative alternative") %>%
filter(sigNew == "Alt_Yes_sig_FALSE" | sigNew == "Alt_Yes_sig_TRUE")
ggplot(df, aes(x = agID2, y = n, color = sigNew, fill = sigNew, label = n)) +
geom_col(position = position_dodge(preserve = 'single')) +
theme_pub() +
theme(legend.position = "top") +
scale_fill_nice_br_b() +
scale_color_nice_br_d() +
geom_text(position = position_dodge(width = 0.8), angle = 45, size = 3) +
labs(
title = "Number of alternative AGs per position",
x = "Alternative AG",
y = "Number of alternative AGs",
color = "Significant event",
fill = "Significant event"
)
Show code
ggsave("features_maxEnt_bar_v2.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")4 Branchpoints at 3’ splice sites
Here we use branchpointer to predict branchpoints for each canonical splice sites (AG). The prediction works only in a defined window from -18 to -44nt from the 3’ splice site.
Note that for each event only one search frame is spanned, which is based on the canonical AG. As an alternative approach, a separate window could be spanned for each AG.
Show code
grBpPacBio = readRDS(file = "./data/grBpPacBioBp.rds")
grBpPacBio$exonID = grBpPacBio$id
grBpPacBio = grBpPacBio[grBpPacBio$exonID %in% ss3PacBioEnd$exonID]
idx = match(grBpPacBio$exonID, ss3PacBioEnd$exonID)
grBpPacBio$regulation = ss3PacBioEnd$regulation[idx]
grBpPacBio$sig = ss3PacBioEnd$sig[idx]
grBpPacBio$sigBp = ifelse(grBpPacBio$branchpoint_prob > 0.52, "Yes", "No")
grBpPacBio$bpID = paste0("BP",names(grBpPacBio), "_", seq_along(grBpPacBio))
names(grBpPacBio) = grBpPacBio$bpID4.1 Branchpoint integration with splice sites
For most canonical 3’ splice sites branchpointer predicts one or two branchpoints (above branchpoint significance probability >0.52). We keep the best, and if present the second best, prediction and assign them to the canonical splice site.
Show code
filterBp = mcols(grBpPacBio) %>%
as.data.frame() %>%
select(exonID, regulation, sig, sigBp, bpID, to_3prime_point, branchpoint_prob, U2_binding_energy, id) %>%
filter(branchpoint_prob > 0.52) %>%
group_by(exonID) %>%
arrange(exonID, desc(branchpoint_prob)) %>%
summarize(exonID, regulation, sig, sigBp, bpID, to_3prime_point, branchpoint_prob, U2_binding_energy, id, bp_order = seq_along(id))
bp1 = filterBp %>% filter(bp_order == 1) # strongest bp
bp2 = filterBp %>% filter(bp_order == 2) # second strongest bp
grAgPacBio$bp1_prob = bp1$branchpoint_prob[match(grAgPacBio$exonID, bp1$exonID)]
grAgPacBio$bp1_position = bp1$to_3prime_point[match(grAgPacBio$exonID, bp1$exonID)]
grAgPacBio$bp1_U2energy = bp1$U2_binding_energy[match(grAgPacBio$exonID, bp1$exonID)]
grAgPacBio$bp1_bpID = bp1$bpID[match(grAgPacBio$exonID, bp1$exonID)]
grAgPacBio$bp1_sig = bp1$sigBp[match(grAgPacBio$exonID, bp1$exonID)]
grAgPacBio$bp2_prob = bp2$branchpoint_prob[match(grAgPacBio$exonID, bp2$exonID)]
grAgPacBio$bp2_position = bp2$to_3prime_point[match(grAgPacBio$exonID, bp2$exonID)]
grAgPacBio$bp2_U2energy = bp2$U2_binding_energy[match(grAgPacBio$exonID, bp2$exonID)]
grAgPacBio$bp2_bpID = bp2$bpID[match(grAgPacBio$exonID, bp2$exonID)]
grAgPacBio$bp2_sig = bp2$sigBp[match(grAgPacBio$exonID, bp2$exonID)]
countVar = subset(grAgPacBio, grAgPacBio$regulation == "shorter_introns" & grAgPacBio$bp1_prob > 0.52)In total 2,831 events have at least one branchpoint above that threshold.
Show code
idx = match(bp2$exonID, bp1$exonID)
d1 = bp1[idx,]
d2 = bp2
d = table(d1$to_3prime_point > d2$to_3prime_point) %>%
as.data.frame() %>%
mutate(s = sum(Freq)) %>%
mutate(per = Freq / s) %>%
mutate(perNice = paste0(round(per, digits = 3) * 100,"%")) %>%
head(1) In 59.5% of the cases the stronger branchpoint is closer to the canonical (1,351 from 2,269). This means that in these 59.5% cases the first branchpoint is also the strongest.
Show code
df = grAgPacBio %>%
as.data.frame() %>%
filter(regulation == "shorter_introns") %>%
mutate(agDist = abs(agPosition - 99)) %>%
filter(!agDist %in% c(3,4,5,6)) %>%
select(agID2, sig, agDist, bp1_position, bp2_position) %>%
filter(agID2 %in% c(0,1,2))
df1 = df %>%
filter(agID2 == 0) %>%
pivot_longer(-c(agID2, sig)) %>%
filter(name %in% c("bp1_position", "bp2_position")) %>%
select(sig,name,value)
df2 = df %>%
pivot_longer(-c(agID2, sig)) %>%
filter(name %in% c("agDist")) %>%
mutate(name = paste0(name, "_", agID2)) %>%
filter(agID2 %in% c(1,2)) %>%
select(sig,name,value)
dfPlot = rbind.data.frame(df1,df2) %>%
filter(value <= 50) %>%
mutate(value = value * -1)
nC = dfPlot %>%
group_by(sig, name) %>%
summarise(count = n()) %>%
mutate(count = myFormat(count))
statTest = dfPlot %>%
group_by(name) %>%
t_test(value ~ sig) %>%
adjust_pvalue(method = "BH") %>%
add_significance("p.adj") %>%
add_xy_position(x = "name", dodge = 0.8)
labNames = c("AG1", "AG2", "BP1", "BP2")
ggplot() +
annotate(geom = "rect", ymin = 0-21, ymax = 0-12, xmin = -Inf, xmax = Inf, color = NA, fill = "grey", alpha = 0.5) +
geom_boxplot(data = dfPlot, aes(x = name, y = value, fill = sig, color = sig)) +
coord_flip() +
scale_x_discrete(labels = labNames) +
ylim(-50,5) +
annotate(geom = "rect", xmin = 0.2, xmax = 0.35, ymin = -50, ymax = 0) +
annotate(geom = "rect", xmin = 0.1, xmax = 0.5, ymin = 0, ymax = Inf) +
theme_nice() +
theme(legend.position = "top") +
scale_color_nice_gr_d() +
scale_fill_nice_gr_b() +
labs(
title = "Alternative AG and branchpoint distances relative to canonical 3'SS",
subtitle = "BP1 = strongest BP, BP2 = second strongest BP",
y = "Distance relative to 3'SS",
x = "Selected features",
color = "Significant event",
fill = "Significant event"
) +
geom_text(data = nC, aes(x = name, y = 5, label = count, group = sig), position = position_dodge(width = 0.8), angle = 45, size = 3) +
stat_pvalue_manual(data = statTest, label = "p", tip.length = 0.02, size = 3) 
Show code
ggsave("maxEnd_Bp_map_v1.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")Show code
df = grAgPacBio %>%
as.data.frame() %>%
filter(regulation == "shorter_introns") %>%
mutate(agDist = abs(agPosition - 99)) %>%
filter(!agDist %in% c(3,4,5,6)) %>%
select(agID2, sig, agDist, bp1_position, bp2_position, isAlternative) %>%
filter(agID2 %in% c(0,1,2))
df1 = df %>%
filter(agID2 == 0) %>%
pivot_longer(-c(agID2, sig, isAlternative)) %>%
filter(name %in% c("bp1_position", "bp2_position")) %>%
select(sig,name,value,isAlternative)
df2 = df %>%
pivot_longer(-c(agID2, sig, isAlternative)) %>%
filter(name %in% c("agDist")) %>%
mutate(name = paste0(name, "_", agID2)) %>%
filter(agID2 %in% c(1,2)) %>%
select(sig,name,value,isAlternative)
df22 = df2 %>% filter(isAlternative == "Yes") %>%
mutate(name = "Alt")
df2 = rbind.data.frame(df2, df22) %>% as.data.frame()
dfPlot = rbind.data.frame(df1,df2) %>%
filter(value <= 50) %>%
mutate(value = value * -1) %>%
mutate(name = factor(name, levels = c("Alt", "agDist_1", "agDist_2", "bp1_position", "bp2_position"))) %>%
mutate(color = ifelse(name != "Alt" & sig == FALSE, "Evt not sig + Both AG",
ifelse(name != "Alt" & sig == TRUE, "Evt sig + Both AG",
ifelse(name == "Alt" & sig == TRUE, "Evt sig + Alt AG", "Evt not sig + Alt AG")))) %>%
mutate(color = factor(color, levels = c("Evt not sig + Both AG", "Evt sig + Both AG", "Evt not sig + Alt AG", "Evt sig + Alt AG"))) %>%
arrange(color)
nC = dfPlot %>%
group_by(color, name) %>%
summarise(count = n()) %>%
mutate(count = myFormat(count))
statTest = dfPlot %>%
group_by(name) %>%
t_test(value ~ color) %>%
adjust_pvalue(method = "BH") %>%
add_significance("p.adj") %>%
add_xy_position(x = "name", dodge = 0.8)
labNames = c("AG-Alt", "AG-1", "AG-2", "BP-1", "BP-2")
ggplot() +
annotate(geom = "rect", ymin = 0-21, ymax = 0-12, xmin = -Inf, xmax = Inf, color = NA, fill = "grey", alpha = 0.5) +
geom_boxplot(data = dfPlot, aes(x = name, y = value, fill = color, color = color)) +
coord_flip() +
scale_x_discrete(labels = labNames) +
ylim(-50,5) +
annotate(geom = "rect", xmin = 0.2, xmax = 0.35, ymin = -50, ymax = 0) +
annotate(geom = "rect", xmin = 0.1, xmax = 0.5, ymin = 0, ymax = Inf) +
theme_nice() +
theme(legend.position = "top") +
scale_color_nice_full_d() +
scale_fill_nice_full_b() +
labs(
title = "Alternative AG and branchpoint distances relative to canonical 3'SS",
subtitle = "BP1 = strongest BP, BP2 = second strongest BP",
y = "Distance relative to 3'SS",
x = "Selected features",
color = "Significant event",
fill = "Significant event"
) +
geom_text(data = nC, aes(x = name, y = 5, label = count, group = color), position = position_dodge(width = 0.8), angle = 45, size = 3) +
stat_pvalue_manual(data = statTest, label = "p", tip.length = 0.02, size = 3) +
guides(fill=guide_legend(ncol=2), color=guide_legend(ncol=2))
Show code
ggsave("maxEnd_Bp_map_v2.pdf", width = 4, height = 4, device = "pdf", path = "./plots/")5 Session Information
Show code
sessionInfo()R version 4.2.1 (2022-06-23)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Big Sur ... 10.16
Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/4.2/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.2/Resources/lib/libRlapack.dylib
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats4 stats graphics grDevices utils datasets methods
[8] base
other attached packages:
[1] ggpubr_0.6.0 rstatix_0.7.2 patchwork_1.1.2
[4] ggrastr_1.0.2 ggsci_3.0.0 tidyr_1.3.0
[7] dplyr_1.1.2 ggplot2_3.4.2 GenomicFeatures_1.49.7
[10] AnnotationDbi_1.59.1 Biobase_2.57.1 rtracklayer_1.57.0
[13] GenomicRanges_1.49.1 GenomeInfoDb_1.33.10 IRanges_2.31.2
[16] S4Vectors_0.35.4 BiocGenerics_0.43.4
loaded via a namespace (and not attached):
[1] bitops_1.0-7 matrixStats_1.0.0
[3] bit64_4.0.5 filelock_1.0.2
[5] progress_1.2.2 httr_1.4.6
[7] tools_4.2.1 backports_1.4.1
[9] utf8_1.2.3 R6_2.5.1
[11] vipor_0.4.5 DBI_1.1.3
[13] colorspace_2.1-0 withr_2.5.0
[15] tidyselect_1.2.0 prettyunits_1.1.1
[17] bit_4.0.5 curl_5.0.1
[19] compiler_4.2.1 textshaping_0.3.6
[21] cli_3.6.1 Cairo_1.6-0
[23] xml2_1.3.4 DelayedArray_0.23.2
[25] labeling_0.4.2 scales_1.2.1
[27] rappdirs_0.3.3 systemfonts_1.0.4
[29] stringr_1.5.0 digest_0.6.31
[31] Rsamtools_2.13.4 rmarkdown_2.22
[33] XVector_0.37.1 pkgconfig_2.0.3
[35] htmltools_0.5.5 MatrixGenerics_1.9.1
[37] dbplyr_2.3.2 fastmap_1.1.1
[39] htmlwidgets_1.6.2 rlang_1.1.1
[41] rstudioapi_0.14 RSQLite_2.3.1
[43] farver_2.1.1 BiocIO_1.7.1
[45] generics_0.1.3 jsonlite_1.8.5
[47] BiocParallel_1.31.13 car_3.1-2
[49] RCurl_1.98-1.12 magrittr_2.0.3
[51] GenomeInfoDbData_1.2.9 Matrix_1.5-4.1
[53] Rcpp_1.0.10 ggbeeswarm_0.7.2
[55] munsell_0.5.0 fansi_1.0.4
[57] ggnewscale_0.4.9 abind_1.4-5
[59] lifecycle_1.0.3 stringi_1.7.12
[61] yaml_2.3.7 carData_3.0-5
[63] SummarizedExperiment_1.27.3 zlibbioc_1.43.0
[65] BiocFileCache_2.5.2 grid_4.2.1
[67] blob_1.2.4 parallel_4.2.1
[69] crayon_1.5.2 lattice_0.21-8
[71] Biostrings_2.65.6 hms_1.1.3
[73] KEGGREST_1.37.3 knitr_1.43
[75] pillar_1.9.0 rjson_0.2.21
[77] ggsignif_0.6.4 codetools_0.2-19
[79] biomaRt_2.53.3 XML_3.99-0.14
[81] glue_1.6.2 evaluate_0.21
[83] png_0.1-8 vctrs_0.6.3
[85] gtable_0.3.3 purrr_1.0.1
[87] cachem_1.0.8 xfun_0.39
[89] broom_1.0.5 restfulr_0.0.15
[91] ragg_1.2.5 tibble_3.2.1
[93] GenomicAlignments_1.33.1 beeswarm_0.4.0
[95] memoise_2.0.1