-
Notifications
You must be signed in to change notification settings - Fork 6
/
plasma-nfl-reanalysis.R
217 lines (187 loc) · 9.48 KB
/
plasma-nfl-reanalysis.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
options(stringsAsFactors = F)
if (interactive()) {
setwd('~/j/cureffi/media/2020/08/')
}
library(reshape2)
library(beeswarm)
# constants
slowcol = '#1B9E77'
fastcol = '#7570B3'
linecol = '#FF0143'
# functions
percent = function(proportion,digits=2) {
return ( gsub(' ','',paste(formatC(proportion*100, digits=digits, format='fg'),"%",sep="") ) )
}
# raw data in the weird format enguage digitizer extracts it in
raw = read.table('thompson-2020-fig4b-raw.csv',sep=',',header=T)
# melt to long format
nfldata = melt(data=raw, id.vars='x', na.rm=T)
colnames(nfldata) = c('years','indiv','nfl')
nfldata$type = substr(nfldata$indiv,1,4)
write.table(nfldata, 'thompson-2020-fig4b.tsv', sep='\t', row.names=F, col.names=T, quote=F)
# other data from table 1
# healthy control distribution properties
hc_percentiles = data.frame(quantile=c(0.0, 0.25, 0.5,0.75,0.90,1.0),
disp=c('min','25th','50th','75th','90th','max'),
nfl=c(2.42, 4.6, 5.81, 7.92, 11.3, 23.7))
# remained-asymptomatic distribution properties
as_percentiles = data.frame(quantile=c(0.0, 0.25, 0.5,0.75,1.0),
disp=c('min','25th','50th','75th','max'),
nfl=c(2.64, 3.5, 5.24, 8.38, 28.35))
# log-linear model of year-on-year increase prior to onset
m = lm(log(nfl) ~ years + indiv, data = subset(nfldata, years < 0))
model_results = as.data.frame(summary(m)$coefficients)
colnames(model_results)[c(1,4)] = c('beta','p')
model_results$expbeta = exp(model_results[,'beta'])
model_results = signif(model_results,3)
model_results[1:2,c('beta','expbeta','p')]
# plot version of 4b with slow and rapid separated
png('~/j/cureffi/media/2020/08/nfl-slow-v-fast.png',width=800,height=400,res=100)
par(mfrow=c(1,2), mar=c(4,4,1,4))
xlims=c(-8,.1)
ylims=c(0,30)
plot(NA, NA, xlim=xlims, ylim=ylims, axes=F, ann=F, xaxs='i', yaxs='i')
axis(side=1, at=seq(-8,0,2), labels=seq(8,0,-2))
mtext(side=1, line=2.5, text='years until onset')
axis(side=2, at=0:4*10, las=2)
mtext(side=2, line=2.5, text='plasma NfL (pg/mL)')
mtext(side=3, line=0, text='slow mutations', font=2, col=slowcol)
for (indiv in unique(nfldata$indiv[nfldata$type=='slow'])) {
subs = nfldata[nfldata$indiv == indiv & nfldata$years < 0,]
points(x=subs$years, y=subs$nfl, pch=20, col=slowcol)
points(x=subs$years, y=subs$nfl, type='l', lwd=2, col=slowcol)
}
abline(h=hc_percentiles$nfl,col=linecol,lwd=0.25)
mtext(side=4, at=hc_percentiles$nfl, col=linecol,las=2, line=0.25, text=hc_percentiles$disp, cex=0.9)
par(xpd=T)
text(x=2.0,y=mean(ylims),label='percentile of healthy controls',col=linecol,srt=270)
par(xpd=F)
xlims=c(-8,.1)
ylims=c(0,30)
plot(NA, NA, xlim=xlims, ylim=ylims, axes=F, ann=F, xaxs='i', yaxs='i')
axis(side=1, at=seq(-8,0,2), labels=seq(8,0,-2))
mtext(side=1, line=2.5, text='years until onset')
axis(side=2, at=0:4*10, las=2)
mtext(side=2, line=2.5, text='plasma NfL (pg/mL)')
mtext(side=3, line=0, text='rapid mutations', font=2, col=fastcol)
for (indiv in unique(nfldata$indiv[nfldata$type=='fast'])) {
subs = nfldata[nfldata$indiv == indiv & nfldata$years < 0,]
points(x=subs$years, y=subs$nfl, pch=20, col=fastcol)
points(x=subs$years, y=subs$nfl, type='l', lwd=2, col=fastcol)
}
abline(h=hc_percentiles$nfl,col=linecol,lwd=0.25)
mtext(side=4, at=hc_percentiles$nfl, col=linecol,las=2, line=0.25, text=hc_percentiles$disp, cex=0.9)
par(xpd=T)
text(x=2.0,y=mean(ylims),label='percentile of healthy controls',col=linecol,srt=270)
par(xpd=F)
dev.off()
# how many converters had a reading >90%tile of healthy controls at any point?
sum(nfldata$nfl[nfldata$years < 0] > hc_percentiles$nfl[hc_percentiles$quantile==.9]) # 14
# how many were within 2 years of onset
sum(nfldata$nfl[nfldata$years < 0 & nfldata$years > -2] > hc_percentiles$nfl[hc_percentiles$quantile==.9]) # 9
# 2nd-highest HC point appears to be about 16.9 pg/mL
hc_2nd_highest = 16.9
known_hc_points = c(hc_percentiles$nfl, hc_2nd_highest)
known_as_points = as_percentiles$nfl
# code used for trial and error to find distributions that match published properties of the actual distributions
# # simulate hc distribution
# known_hc_points = c(hc_percentiles$nfl)
# for (i in 1:10000) {
# set.seed(i)
# m = 5.85
# s = 1.7
# hc = sort(c(10^(rnorm(mean = log10(m), sd=log10(s), n=45-length(known_hc_points))), known_hc_points))
# hc = pmax(hc, 2.42) # llq?
# qtiles = quantile(x=hc, probs = hc_percentiles$quantile)
#
# if (sum(abs(as.numeric(qtiles) - hc_percentiles$nfl)) < .01) {
# cat(i)
# flush.console()
# break
# }
# }
# by some trial and error found this distribution comes very close - all percentiles match within < 0.1
seedno = 3204
m = 5.85
s = 1.7
set.seed(seedno)
hc = sort(c(10^(rnorm(mean = log10(m), sd=log10(s), n=45-length(known_hc_points))), known_hc_points))
hc = pmax(hc, 2.42) # llq?
hc_qtiles = quantile(x=hc, probs = hc_percentiles$quantile)
# and by more trial and error found this distribution comes very close - all percentiles match within < 0.1
seedno=2042
set.seed(seedno)
m = 5.7
s = 1.7
as = sort(c(10^(rnorm(mean = log10(m), sd=log10(s), n=70-length(known_as_points))), known_as_points))
as = pmax(as, 2.64) # this appears to be llq or close to it
# if you don't put a lower limit, then increasing the variance skews too low before the 75th percentile hits 8+ pg/mL
as_qtiles = quantile(x=as, probs = as_percentiles$quantile)
all_asymp = nfldata[nfldata$years < 0,c('nfl','years')]
all_asymp$onset2y = all_asymp$years > -2
all_asymp = all_asymp[,-2] # remove years col
all_asymp = rbind(all_asymp, data.frame(nfl=as, onset2y=rep(F,length(as))))
all_asymp$x = 2
all_points = rbind(all_asymp, data.frame(nfl=hc, onset2y=rep(F,length(hc)), x=rep(1, length(hc))))
onset2y_color = '#CD00CD'
noonset_color = '#383838'
all_points$color[all_points$onset2y] = onset2y_color
all_points$color[!all_points$onset2y] = noonset_color
png('cross-sectional-simulation.png',width=800,height=500,res=100)
par(mar=c(4,4,2,8))
beeswarm(nfl ~ x, data=all_points, pch=19, pwcol=all_points$color,
axes=F, xaxs='i', yaxs='i', xlab='', ylab='', ylim=c(0,30))
axis(side=1, labels=NA, lwd.ticks=0)
axis(side=2, at=0:6*5, las=2)
mtext(side=1, at=c(1,2), text=c('healthy\ncontrols','asymptomatic-at-the-time\ncarriers'), line=1)
mtext(side=2, line=2.5, text='plasma NfL (pg/mL)')
mtext(side=3, line=0.5, text='cross-sectional analysis', col='#777777', font=1)
mtext(side=3, line=-0.5, text='SIMULATION ONLY', col='#777777', font=2)
abline(h=hc_percentiles$nfl[hc_percentiles$quantile==.9], col='red')
mtext(side=4, line=0.25, at=hc_percentiles$nfl[hc_percentiles$quantile==.9], las=2, col='red', text='90th\npercentile\ncontrol\ncutoff')
par(xpd=T)
legend(x=2.25, y=29, c('<2 years from onset','>2 years from onset'), col=c(onset2y_color, noonset_color), pch=19)
par(xpd=F)
dev.off()
# what proportion of symptomatic IPD in the study were rapid v slow?
symptomatic_genotypes = data.frame(mutation=c('E200K','4-OPRI','E196K','V210I','D178N','P102L','A117V','P105L','5-OPRI','6-OPRI','Q212P','Y163X'),
pace=c(rep('rapid',5),rep('slow',7)),
n=c(10,3,1,1,6,26,5,1,7,19,1,3))
sum(symptomatic_genotypes$n[symptomatic_genotypes$pace=='rapid'])
sum(symptomatic_genotypes$n[symptomatic_genotypes$pace=='slow'])
# make table of duration and prevalence
# high-penetrance mutation prevalence from Minikel 2019
prevalence = read.table('https://raw.githubusercontent.com/ericminikel/prnp_onset/master/data/mutation_prevalence.tsv',sep='\t',header=T)
# UK surveillance center data from Minikel 2016
cases_raw = t(read.table('https://raw.githubusercontent.com/ericminikel/prnp_penetrance/master/supplement/table_s01_surveillance_allele_counts.tsv',header=F,sep='\t'))
uk_cases = as.data.frame(cases_raw[9:68,c(1,9)])
colnames(uk_cases) = c('variant','count')
uk_cases$count = as.integer(uk_cases$count)
uk_cases$count[is.na(uk_cases$count)] = 0
prevalence$uk_cases = uk_cases$count[match(prevalence$variant, uk_cases$variant)]
prevalence$uk_prev = prevalence$uk_cases / sum(prevalence$uk_cases)
mutation_tbl = prevalence[1:7,c('rank','variant','proportion','uk_prev')]
mutation_tbl$median_duration_months = as.integer(NA)
mutation_tbl$surv24mo = as.numeric(NA)
for (i in 1:nrow(mutation_tbl)) {
duration = read.table(paste0('https://raw.githubusercontent.com/ericminikel/prnp_onset/master/data/duration_',mutation_tbl$variant[i],'.tsv'),sep='\t',header=T)
# see https://github.com/ericminikel/prnp_onset/blob/master/src/shared_code.R#L73-L83 for how to calculate median, adapted here:
duration$lt1 = duration$alive - duration$censored/2
duration$qt = duration$dead/duration$lt1
duration$pt = 0.0 # Pt = cumulative survival
duration$pt[1] = 1.0
for (j in 2:nrow(duration)) {
duration$pt[j] = (1 - duration$qt[j]) * duration$pt[j-1]
}
mutation_tbl$median_duration_months[i] = min(duration$months[duration$pt <= 0.5], na.rm=T)
mutation_tbl$surv24mo[i] = duration$pt[duration$months==24]
}
mutation_tbl$world_pct = percent(mutation_tbl$proportion,digits=0)
mutation_tbl$uk_pct = percent(round(mutation_tbl$uk_prev,2),digits=0)
mutation_tbl$median_duration = paste0(mutation_tbl$median_duration_months, ' months')
mutation_tbl$median_duration[mutation_tbl$median_duration_months > 12] = paste0(round(mutation_tbl$median_duration_months[mutation_tbl$median_duration_months > 12]/12,1), ' years')
# proportion surviving after 24 mo, for body of post:
sum(mutation_tbl$proportion * mutation_tbl$surv24mo)
sum(mutation_tbl$uk_prev * mutation_tbl$surv24mo)
# table for post
mutation_tbl[,c('variant','median_duration','world_pct','uk_pct')]