Table II. Effectiveness of Protein Prospector metrics as classifiers of target versus decoy cross-link matches.
Parameter | r coefa | # Targetb | Specificity | Threshold |
---|---|---|---|---|
SVM dvalc | 0.581 | 2349 | 0.927 | 0.0 |
scorediffd | 0.544 | 2258 | 0.923 | 8.5 |
pep2.pExpe | 0.441 | 1405 | 0.926 | −0.2 |
XL pExpf | 0.439 | 1045 | 0.926 | 9.9 |
pep2.score | 0.439 | 1621 | 0.925 | 13.6 |
Low score | 0.430 | 1785 | 0.924 | 15.9 |
XL score | 0.408 | 1122 | 0.926 | 58.3 |
% TIC matchedg | 0.390 | 1092 | 0.926 | 79.6% |
pep2.norm_scoreh | 0.297 | 733 | 0.926 | 2.4 |
pep1.pExp | 0.280 | 425 | 0.924 | 9.2 |
pep1.score | 0.275 | 532 | 0.925 | 55.3 |
pep1.norm_score | 0.143 | 342 | 0.924 | 4.7 |
pep1.lengthi | 0.104 | 375 | 0.926 | 24 |
ppm | 0.086 | NA | NA | NA |
z | 0.067 | NA | NA | NA |
mz | 0.004 | NA | NA | NA |
pep2.length | −0.024 | NA | NA | NA |
pep1.rankj | −0.204 | NA | NA | NA |
pep2.rank | −0.308 | NA | NA | NA |
a Point biserial correlation coefficient between Prospector metric and matches to the target database.
b Number of spectral matches classified as positive cross-link hits at the given score threshold, chosen to achieve equal specificity (see text).
c Score of final SVM classifier. Model was trained as described in the text; reported here is the result of the final classification.
d Difference in score between the top cross-linked match and the top linear match.
e −log10(Exppep2), where Exppep2 + expectation value of weaker peptide.
f −log10(ExpXL), where XL refers to the complete cross-link.
g Percentage of total ion current intensity in the peaklist that can be explained by the cross-link.
h Score normalized by length of peptide.
i Length of peptide in amino acids.
j Rank of individual peptide match (e.g. rank 1 indicates top match to spectrum).