Database Reference
In-Depth Information
4
Expected log-ratio: 5:1
40
5
P
1
I
i
30
3
4
P
i
3
20
N
i
2
2
P
2
10
1
1
0
1
Heavy Isotopologue Intensity (×10
5
)
2
3
-4
0
Peptide Log-Ratio
-2
24
Figure 8.13
(See color insert following page 224.) Peptide abundance ratio
and signal-to-noise ratio estimation via PCA (left). Signal-to-noise ratio is
inversely correlated with the variability and bias of peptide abundance ratio
(right).
of processors used (up to 32) after less than 10% overhead as described in
Section 8.6.1.1.
8.6.1.3
Protein Abundance Ratio Estimation with Confidence
Interval Evaluation
To evaluate protein abundance ratios from quantitative proteomics measure-
ments, two types of statistical estimation should be employed:
point estima-
tion
and
interval estimation
. The point estimation gives an abundance ratio
for every quantified protein, which best approximates the true abundance ra-
tio. Unfortunately, the point estimation provides no information about protein
quantification precision, which can significantly vary across different proteins.
Generally, a protein should have better quantification precision if it has more
peptides quantified from mass spectral data of higher signal-to-noise ratio.
It is misleading in quantitative proteomics to treat all proteins' abundance
ratios identically, regardless of their estimation precision.
The interval estimation complements the point estimation by providing con-
fidence intervals for protein abundance ratios. If 90% of quantified proteins
have confidence intervals that contain their true abundance ratios, then con-
fidence intervals are estimated at a 90% confidence level. The confidence level
for the interval estimation in quantitative proteomics is analogous to the true
positive rate for protein identification in qualitative proteomics. More impor-
tantly, at a given confidence level, the confidence interval intuitively reflects
the quantification precision for each protein as an error bar of the abundance
ratio estimate.
