R Notebook

This notebook takes the output for the MAF Pipeline (mycsvfile.csv) and does:

Calculates true pos, false pos, and false neg and makes a precision recall curve for all tools
Finds total numbers of variants found by tools at different allele frequencies and sequencing depths
Looks at correlation between SNPS in the golden vcf and worflow vcf

#knitr::opts_knit$set(root.dir = "/Users/marissaknoll/Desktop/MAF/")
#run in working directory
#setwd("/Users/marissaknoll/Desktop/MAF/")

#library(available)
#available("Vivaldi", browse = FALSE)

Loading Libraries

library('plyr')
library('tidyverse')

-- Attaching packages --------------------------------------- tidyverse 1.3.0 --

v ggplot2 3.3.2     v purrr   0.3.3
v tibble  3.0.4     v dplyr   1.0.2
v tidyr   1.0.0     v stringr 1.4.0
v readr   1.3.1     v forcats 0.4.0

-- Conflicts ------------------------------------------ tidyverse_conflicts() --
x dplyr::arrange()   masks plyr::arrange()
x purrr::compact()   masks plyr::compact()
x dplyr::count()     masks plyr::count()
x dplyr::failwith()  masks plyr::failwith()
x dplyr::filter()    masks stats::filter()
x dplyr::id()        masks plyr::id()
x dplyr::lag()       masks stats::lag()
x dplyr::mutate()    masks plyr::mutate()
x dplyr::rename()    masks plyr::rename()
x dplyr::summarise() masks plyr::summarise()
x dplyr::summarize() masks plyr::summarize()

library("plotrix") 
library('ggplot2')
library('ggpubr')


Attaching package: 'ggpubr'

The following object is masked from 'package:plyr':

    mutate

library("MLmetrics")


Attaching package: 'MLmetrics'

The following object is masked from 'package:base':

    Recall

library("RColorBrewer")
        
MyColors = c("#e41a1c", "#377eb8", "#4daf4a","#984ea3", "#ff6600", "#ffcc33", "#f781bf", "#a65628", "#999999", "#006600", "#993333", "#cc99ff", "#663300")
# This is basically RColorBrewer Set1 but I darkened yellow to make it easier to see then changed orange a bit to differ more from yellow
#
SEGMENTLIST = c("H1N1_PB2","H1N1_PB1","H1N1_PA","H1N1_HA","H1N1_NP","H1N1_NA","H1N1_MP","H1N1_NS")
ggplot2::theme_set(theme_minimal())

totalbases = 29903 # at some point should change this from being hardcoded to make this more flexible
readsim_cov = 100000 #hardocding this for now so my code runs

Reading in Variant Data and Evaluating

AV = read.csv(af_report,header = T)
#make the segments in a specific order
AV$chrom = factor(AV$chrom, levels = SEGMENTLIST)

AV$sample_id = as.character(AV$sample_id)
AV$dp = as.numeric(AV$dp)
AV = separate(AV, sample_id, sep = "_", 
               into = c("strain","AF","allele_freq","frac","seq_depth","tool","parameter","parameter2")) %>%
  select(allele_freq,seq_depth,tool, parameter, chrom,pos,af_golden,af_workflow,ref,alt,dp) %>%
  filter(ref == "A" | ref == "C" | ref == "T" | ref == "G") %>% 
  filter(alt == "A" | alt == "C" | alt == "T" | alt == "G") %>% droplevels()

Warning: Expected 8 pieces. Missing pieces filled with `NA` in 65999 rows [1, 2,
3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...].

AV$seq_depth = as.numeric(AV$seq_depth)
AV$coverage = (AV$seq_depth * readsim_cov) #variable called in pipeline, corresponds to highest coverage
AV$coverage = as.character(AV$coverage)
head(AV)

ABCDEFGHIJ0123456789

	allele_freq <chr>	seq_depth <dbl>	tool <chr>	parameter <chr>	chrom <fct>	pos <int>	af_golden <dbl>	ref <fct>
1	0.03	0.005	ivar	NA	H1N1_HA	1644	0.03	G
2	0.03	0.005	ivar	NA	H1N1_MP	862	0.03	T
3	0.03	0.005	ivar	NA	H1N1_MP	875	0.03	C
4	0.03	0.005	ivar	NA	H1N1_NA	1289	0.03	G
5	0.03	0.005	ivar	NA	H1N1_NA	1371	0.03	G
6	0.03	0.005	ivar	NA	H1N1_NP	1392	0.03	C

variant_pos = ggplot(AV, aes(x = pos, y = tool)) +
  geom_point() +
  facet_grid(~chrom, scales = "free_x") + theme_minimal(base_size = 6)
print(variant_pos)

Reading in Golden Data and Evaluating

viral_golden = read.table(golden_vcf, stringsAsFactors = F)
colnames(viral_golden) = c("chrom","pos","ID","ref","alt","qual","filter","info","codes","genotype")
viral_golden = separate(viral_golden, info, sep = "=", into = c(NA,"AF"))
viral_golden$chrom = factor(viral_golden$chrom, levels = SEGMENTLIST)
head(viral_golden)

ABCDEFGHIJ0123456789

	chrom <fct>	pos <int>	ID <chr>	ref <chr>	alt <chr>	qual <chr>	filter <chr>	AF <chr>	codes <chr>
1	H1N1_HA	382	.	T	C	.	.	0.03	GT:AD:DP:GQ:PL
2	H1N1_HA	439	.	A	C	.	.	0.03	GT:AD:DP:GQ:PL
3	H1N1_HA	930	.	G	A	.	.	0.03	GT:AD:DP:GQ:PL
4	H1N1_HA	976	.	A	G	.	.	0.03	GT:AD:DP:GQ:PL
5	H1N1_HA	1055	.	G	A	.	.	0.03	GT:AD:DP:GQ:PL
6	H1N1_HA	1070	.	G	A	.	.	0.03	GT:AD:DP:GQ:PL

dim(viral_golden)

[1] 111  10

viral_golden_pos = ggplot(viral_golden, aes(x = pos, y = ref, color = alt)) +
  geom_point() +
  facet_grid(~chrom, scales = "free_x")+ theme_minimal(base_size = 6)
print(viral_golden_pos)

SetAF - Separating Data into TP, FP, FN

AV_setAF = filter(AV, allele_freq != "random") %>% droplevels()

AV_setAF_falsepos = filter(AV_setAF, af_golden == 0 & af_workflow != 0) %>% droplevels()
  AV_setAF_falsepos$category = c("FP")
AV_setAF_falseneg = filter(AV_setAF, af_workflow == 0 & af_golden != 0) %>% droplevels()
  AV_setAF_falseneg$category = c("FN")
AV_setAF_truepos = filter(AV_setAF, af_golden != 0 & af_workflow != 0)
  AV_setAF_truepos$category = c("TP")
  
AV_setAF = rbind(AV_setAF_truepos, AV_setAF_falseneg, AV_setAF_falsepos)

FP_All = group_by(AV_setAF_falsepos, allele_freq,coverage,tool) %>% tally()
  colnames(FP_All) = c("allele_freq","coverage","tool","FP")
FN_All = group_by(AV_setAF_falseneg, allele_freq,coverage,tool) %>% tally()
  colnames(FN_All) = c("allele_freq","coverage","tool","FN")
TP_All = group_by(AV_setAF_truepos, allele_freq,coverage,tool) %>% tally()
  colnames(TP_All) = c("allele_freq","coverage","tool","TP")
    
AllVar = merge(FP_All,FN_All, by=c("allele_freq","coverage","tool"),all=T)
AllVar = merge(AllVar,TP_All, by=c("allele_freq","coverage","tool"), all=T)
head(AllVar)

ABCDEFGHIJ0123456789

	allele_freq <chr>	coverage <chr>	tool <chr>	FP <int>	FN <int>	TP <int>
1	0.01	100	filtered	NA	87	24
2	0.01	100	freebayes-default	85	81	30
3	0.01	100	freebayes-pooled-continuous	85	81	30
4	0.01	100	ivar	NA	111	NA
5	0.01	100	lofreq	NA	111	NA
6	0.01	100	mutect2-default	NA	106	5

AllVar[is.na(AllVar)] = 0
head(AllVar)

ABCDEFGHIJ0123456789

	allele_freq <chr>	coverage <chr>	tool <chr>	FP <dbl>	FN <dbl>	TP <dbl>
1	0.01	100	filtered	0	87	24
2	0.01	100	freebayes-default	85	81	30
3	0.01	100	freebayes-pooled-continuous	85	81	30
4	0.01	100	ivar	0	111	0
5	0.01	100	lofreq	0	111	0
6	0.01	100	mutect2-default	0	106	5

AllVar$prec = (AllVar$TP)/(AllVar$TP + AllVar$FP)
AllVar$recall = (AllVar$TP)/(AllVar$TP + AllVar$FN)
AllVar$area = AllVar$prec * AllVar$recall
head(AllVar)

ABCDEFGHIJ0123456789

	allele_freq <chr>	coverage <chr>	tool <chr>	FP <dbl>	FN <dbl>	TP <dbl>	prec <dbl>	recall <dbl>
1	0.01	100	filtered	0	87	24	1.0000000	0.21621622
2	0.01	100	freebayes-default	85	81	30	0.2608696	0.27027027
3	0.01	100	freebayes-pooled-continuous	85	81	30	0.2608696	0.27027027
4	0.01	100	ivar	0	111	0	NaN	0.00000000
5	0.01	100	lofreq	0	111	0	NaN	0.00000000
6	0.01	100	mutect2-default	0	106	5	1.0000000	0.04504505

# SetAF Data
ggplot(AV_setAF, aes(x = tool, fill = category)) +
  geom_bar() +
  facet_grid(allele_freq~coverage) +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1)) +
  scale_fill_brewer(palette = "Dark2")

# plot as percentage of FN as well

RandomAF - Separating Data into TP, FP, FN

AV_randomAF = filter(AV, allele_freq == "random") %>% droplevels()

AV_randomAF_falsepos = filter(AV_randomAF, af_golden == 0 & af_workflow != 0) %>% droplevels()
  AV_randomAF_falsepos$category = c("FP")
AV_randomAF_falseneg = filter(AV_randomAF, af_workflow == 0 & af_golden != 0) %>% droplevels()
  AV_randomAF_falseneg$category = c("FN")
AV_randomAF_truepos = filter(AV_randomAF, af_golden != 0 & af_workflow != 0)
  AV_randomAF_truepos$category = c("TP")
  
AV_randomAF = rbind(AV_randomAF_truepos, AV_randomAF_falseneg, AV_randomAF_falsepos)

rFP_All = group_by(AV_randomAF_falsepos, allele_freq,coverage,tool) %>% tally()
  colnames(rFP_All) = c("allele_freq","coverage","tool","FP")
rFN_All = group_by(AV_randomAF_falseneg, allele_freq,coverage,tool) %>% tally()
  colnames(rFN_All) = c("allele_freq","coverage","tool","FN")
rTP_All = group_by(AV_randomAF_truepos, allele_freq,coverage,tool) %>% tally()
  colnames(rTP_All) = c("allele_freq","coverage","tool","TP")

rAllVar = merge(rFP_All,rFN_All, by=c("allele_freq","coverage","tool"),all=T)
rAllVar = merge(rAllVar, rTP_All, by=c("allele_freq","coverage","tool"), all=T)
head(rAllVar)

ABCDEFGHIJ0123456789

	allele_freq <chr>	coverage <chr>	tool <chr>	FP <int>	FN <int>	TP <int>
1	random	100	filtered	3	4	107
2	random	100	freebayes-default	80	10	101
3	random	100	freebayes-pooled-continuous	80	10	101
4	random	100	ivar	NA	22	89
5	random	100	lofreq	NA	13	98
6	random	100	mutect2-default	NA	56	55

rAllVar[is.na(rAllVar)] = 0

rAllVar$prec = (rAllVar$TP)/(rAllVar$TP + rAllVar$FP)
rAllVar$recall = (rAllVar$TP)/(rAllVar$TP + rAllVar$FN)
rAllVar$area = rAllVar$prec * rAllVar$recall
head(rAllVar)

ABCDEFGHIJ0123456789

	allele_freq <chr>	coverage <chr>	tool <chr>	FP <dbl>	FN <dbl>	TP <dbl>	prec <dbl>	recall <dbl>
1	random	100	filtered	3	4	107	0.9727273	0.9639640
2	random	100	freebayes-default	80	10	101	0.5580110	0.9099099
3	random	100	freebayes-pooled-continuous	80	10	101	0.5580110	0.9099099
4	random	100	ivar	0	22	89	1.0000000	0.8018018
5	random	100	lofreq	0	13	98	1.0000000	0.8828829
6	random	100	mutect2-default	0	56	55	1.0000000	0.4954955

# RandomAF Data
ggplot(AV_randomAF, aes(x = tool, fill = category)) +
  geom_bar() +
  facet_grid(allele_freq~coverage) +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1)) +
  scale_fill_brewer(palette = "Dark2")

# plot as percentage of FN as well

SetAF - Correlation Between Golden and Workflow

#Looking at all SetAF Data
AV_Corr_setAF <- 
  ggplot(AV_setAF, aes(x = af_golden,y = af_workflow, color = category)) +
  geom_point(size = 3, alpha = 0.4) +
  scale_color_manual(values = MyColors) +
  facet_grid(coverage~tool) + 
  geom_smooth(method='lm', size = 1.5) +
  theme(plot.title = element_text(hjust = 0.5))+ theme_minimal(base_size = 6)
print(AV_Corr_setAF)

`geom_smooth()` using formula 'y ~ x'

Warning in qt((1 - level)/2, df): NaNs produced

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

#ggsave(AV_Corr_setAF, file = "AllVarSet_Correlation_Filtered.png",
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")

RandomAF - Correlation Between Golden and Workflow

##Looking at all RandomAF Data
  AV_Corr_randomAF <- 
  ggplot(AV_randomAF, aes(x = af_golden,y = af_workflow, color = category)) +
  geom_point(size = 3, alpha = 0.4) +
  scale_color_manual(values = MyColors) +
  facet_grid(coverage~tool) + 
  geom_smooth(method='lm', size = 1.5) +
  theme(plot.title = element_text(hjust = 0.5))+ theme_minimal(base_size = 6)
print(AV_Corr_randomAF)

`geom_smooth()` using formula 'y ~ x'

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in qt((1 - level)/2, df): NaNs produced

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

Warning in max(ids, na.rm = TRUE): no non-missing arguments to max; returning -
Inf

#ggsave("AllVarRandom_Correlation_Filtered.png",AV_Corr_randomAF,
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")

Stats for SetAF Data

detach(package:plyr, unload=TRUE)
coeff_var <- function(x) {
  CV <- sd(x) / mean(x) * 100
  return(CV)
}

AV_setAF_split = group_by(AV_setAF_truepos, tool, allele_freq, coverage) %>%
  summarize(mean = mean(af_workflow), sd = sd(af_workflow), variance = coeff_var(af_workflow)) %>% droplevels()

`summarise()` regrouping output by 'tool', 'allele_freq' (override with `.groups` argument)

Mean_setAF = ggplot(AV_setAF_split, aes(x = coverage, y = mean, color = allele_freq)) + 
  geom_point() +
  #facet_grid(cols = vars(allele_freq)) +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
print(Mean_setAF)

#ggsave("SetAF_Mean_AlleleFrequency.png",Mean_setAF,
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")
#visulize the effect of downsampling. 
CoeffVar_setAF = ggplot(AV_setAF_split, aes(x = coverage, y = variance, color = allele_freq)) + 
  geom_point() +
  #facet_grid(cols = vars(allele_freq)) +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
print(CoeffVar_setAF)

Warning: Removed 3 rows containing missing values (geom_point).

#ggsave("SetAF_CoeffVar_AlleleFrequency.png",CoeffVar_setAF,
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")
#variances above 10000 are simliar

Stats for RandomAF Data

AV_randomAF_split = group_by(AV_randomAF_truepos, tool, coverage) %>%
  summarize(mean = mean(af_workflow), sd = sd(af_workflow), variance = coeff_var(af_workflow)) %>% droplevels()

`summarise()` regrouping output by 'tool' (override with `.groups` argument)

Mean_randomAF = ggplot(AV_randomAF_split, aes(x = coverage, y = mean, color = tool)) + 
  geom_point() +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
print(Mean_randomAF)

#ggsave("RandomAF_Mean_AlleleFrequency.png",Mean_randomAF,
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")

CoeffVar_randomAF = ggplot(AV_randomAF_split, aes(x = coverage, y = variance, color = tool)) + 
  geom_point() +
  theme(text = element_text(size = 15),
        axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
print(CoeffVar_randomAF)

#ggsave("RandomAF_CoeffVar_AlleleFrequency.png",CoeffVar_randomAF,
#       path = "/Users/marissaknoll/Desktop/MAF/H1N1_Images")

ggplot(AV_setAF, aes(x = log10(dp), y = af_workflow, color = tool)) +
  geom_point() +
  facet_grid(rows = vars(coverage), cols = vars(allele_freq)) +
  xlim(0,6)

Warning: Removed 3210 rows containing missing values (geom_point).

# do this with all bamfiles, not just variants
ggplot(AV_setAF, aes(x = log10(dp), fill = tool)) +
  geom_histogram()+
  facet_grid(rows = vars(coverage), cols = vars(allele_freq))

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Warning: Removed 3210 rows containing non-finite values (stat_bin).

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