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
217
218
219
220
221
222
223
use itertools::Itertools;
use mzcore::{
prelude::MassMode,
quantities::WithinTolerance,
system::{MassOverCharge, Ratio},
};
use mzdata::prelude::*;
use serde::{Deserialize, Serialize};
use crate::{
annotation::model::MatchingParameters, fragment::Fragment, spectrum::AnnotatedSpectrum,
};
impl AnnotatedSpectrum {
/// Get a false discovery rate estimation for this annotation. See the [`Fdr`] struct for all
/// statistics that can be retrieved. The returned tuple has the FDR for all peptides combined
/// as first item and as second item a vector with for each peptide its individual scores.
///
/// # Estimation
/// It estimated the FDR by permutation. It takes all MZs for all fragments (or only the subset
/// for each separate peptide) and counts how many match with the spectrum with offset
/// `-25..=25 + π`. The returned result give insight in the average number of matches and
/// standard deviation. The `π` offset is needed to guarantee non integer offsets preventing
/// spurious matches from 1 Da isotopes.
pub fn fdr(
&self,
fragments: &[Fragment],
parameters: &MatchingParameters,
mass_mode: MassMode,
) -> (Fdr, Vec<Vec<Fdr>>) {
let mzs = fragments
.iter()
.filter_map(|f| {
f.mz(mass_mode)
.map(|mz| (mz, f.peptidoform_ion_index, f.peptidoform_index))
})
.filter(|(mz, _, _)| parameters.mz_range.contains(mz))
.collect_vec();
let individual_peptides = self
.analytes
.iter()
.enumerate()
.filter_map(|a| match &a.1.target {
crate::mzspeclib::AnalyteTarget::PeptidoformIon(pep) => Some((a.0, pep)),
_ => None,
})
.map(|(peptidoform_ion_index, peptidoform)| {
peptidoform
.peptidoforms()
.iter()
.enumerate()
.map(|(peptidoform_index, _)| {
self.internal_fdr(
mzs.iter()
.filter_map(|(mz, pi, ppi)| {
(*pi == Some(peptidoform_ion_index)
&& *ppi == Some(peptidoform_index))
.then_some(*mz)
})
.collect_vec()
.as_slice(),
parameters,
)
})
.collect()
})
.collect();
(
self.internal_fdr(
mzs.iter().map(|(mz, _, _)| *mz).collect_vec().as_slice(),
parameters,
),
individual_peptides,
)
}
fn internal_fdr(&self, mzs: &[MassOverCharge], parameters: &MatchingParameters) -> Fdr {
let mut results = Vec::with_capacity(51);
let total_intensity = self.peaks.iter().map(|s| s.intensity).sum::<f32>();
if self.peaks.is_empty() {
return Fdr::default();
}
for offset in -25..=25 {
let peaks = self
.peaks
.iter()
.map(|p| {
p.mz + MassOverCharge::new::<mzcore::system::thomson>(
std::f64::consts::PI + f64::from(offset),
)
})
.collect_vec();
let mut peak_annotated = vec![false; peaks.len()];
let mut number_peaks_annotated = 0;
let mut intensity_annotated = 0.0;
for mass in mzs {
// Get the index of the element closest to this value (spectrum is defined to always be sorted)
let index = peaks
.binary_search_by(|p| p.value.total_cmp(&mass.value))
.unwrap_or_else(|i| i);
// Check index-1, index and index+1 (if existing) to find the one with the lowest ppm
let mut closest = (0, Ratio::new::<mzcore::system::ratio::ppm>(f64::INFINITY));
#[expect(clippy::needless_range_loop)] // I like this better
for i in if index == 0 { 0 } else { index - 1 }
..=(index + 1).min(self.peaks.len().saturating_sub(1))
{
let ppm = peaks[i].ppm(*mass);
if ppm < closest.1 {
closest = (i, ppm);
}
}
if parameters.tolerance.within(&self.peaks[closest.0].mz, mass)
&& !peak_annotated[closest.0]
{
number_peaks_annotated += 1;
intensity_annotated += self.peaks[closest.0].intensity;
peak_annotated[closest.0] = true;
}
}
results.push((
f64::from(number_peaks_annotated) / self.peaks.len() as f64,
intensity_annotated / total_intensity,
));
}
let peaks_average = results.iter().map(|r| r.0).sum::<f64>() / results.len() as f64;
let peaks_st_dev = (results
.iter()
.map(|x| (x.0 - peaks_average).powi(2))
.sum::<f64>()
/ results.len() as f64)
.sqrt();
let intensity_average = results.iter().map(|r| r.1).sum::<f32>() / results.len() as f32;
let intensity_st_dev = (results
.iter()
.map(|x| (x.1 - intensity_average).powi(2))
.sum::<f32>()
/ results.len() as f32)
.sqrt();
let actual: (u32, f32) = self
.peaks
.iter()
.filter(|p| !p.annotations.is_empty())
.fold((0, 0.0), |acc, p| (acc.0 + 1, acc.1 + p.intensity));
Fdr {
peaks_actual: f64::from(actual.0) / self.peaks.len() as f64,
peaks_average_false: peaks_average,
peaks_standard_deviation_false: peaks_st_dev,
intensity_actual: actual.1 / total_intensity,
intensity_average_false: intensity_average,
intensity_standard_deviation_false: intensity_st_dev,
}
}
}
/// A false discovery rate for an annotation to a spectrum
#[derive(Clone, Copy, Debug, Deserialize, PartialEq, Serialize)]
pub struct Fdr {
/// The fraction of the total (assumed to be true) peaks that could be annotated
pub peaks_actual: f64,
/// The average fraction of the false peaks that could be annotated
pub peaks_average_false: f64,
/// The standard deviation of the false peaks that could be annotated
pub peaks_standard_deviation_false: f64,
/// The fraction of the total (assumed to be true) intensity that could be annotated
pub intensity_actual: f32,
/// The average fraction of the false intensity that could be annotated
pub intensity_average_false: f32,
/// The standard deviation of the false intensity that could be annotated
pub intensity_standard_deviation_false: f32,
}
impl Default for Fdr {
fn default() -> Self {
Self {
peaks_actual: 0.0,
peaks_average_false: 0.0,
peaks_standard_deviation_false: 0.0,
intensity_actual: 0.0,
intensity_average_false: 0.0,
intensity_standard_deviation_false: 0.0,
}
}
}
impl Fdr {
/// Get the false discovery rate (as a fraction).
/// The average number of false peaks annotated divided by the average number of annotated peaks.
pub fn peaks_fdr(self) -> f64 {
self.peaks_average_false / self.peaks_actual
}
/// Get the number of standard deviations the number of annotated peaks is from the average number of false annotations.
pub fn peaks_sigma(self) -> f64 {
(self.peaks_actual - self.peaks_average_false) / self.peaks_standard_deviation_false
}
/// Get the peaks score of this annotation. Defined as the log2 of the sigma.
pub fn peaks_score(self) -> f64 {
self.peaks_sigma().log2()
}
/// Get the false discovery rate (as a fraction).
/// The average number of false intensity annotated divided by the average number of annotated intensity.
pub fn intensity_fdr(self) -> f32 {
self.intensity_average_false / self.intensity_actual
}
/// Get the number of standard deviations the annotated intensity is from the average false annotations.
pub fn intensity_sigma(self) -> f32 {
(self.intensity_actual - self.intensity_average_false)
/ self.intensity_standard_deviation_false
}
/// Get the intensity score of this annotation. Defined as the log2 of the sigma.
pub fn intensity_score(self) -> f32 {
self.intensity_sigma().log2()
}
}