Q5 Algorithm: Sample Results (Ryan Lilien)

Probabilistic Disease Classification of Expression-Dependent Proteomic Data from Mass Spectrometry of Human Serum: Ryan H. Lilien, Hany Farid and Bruce R. Donald; Journal of Computational Biology, 2003; 10(6): 925-946.

A

B

C

Classification Results. The probability classification threshold vs. percent-classified (Classif), percent-correctly classified (Correct), positive predictive value (PPV), sensitivity (Sens), and specificity (Spec) for three D-runs of Q5 on the ovarian cancer datasets. (A) OC-H4, 95% of samples used in training. (B) OC-WCX2a, 95% of samples used in training. (C) OC-WCX2b, 95% of samples used in training. Increased probability classification thresholds increase Q5’s percent-correctly classified, positive predictive value, sensitivity, and specificity while decreasing the percent-classified.

A

B

C
Classification Results. The probability classification threshold vs. percent-classified (Classif), percent-correctly classified (Correct), positive predictive value (PPV), sensitivity (Sens), and specificity (Spec) for three D-runs of Q5 on the ovarian cancer datasets. (A) OC-H4, 95% of samples used in training. (B) OC-WCX2a, 95% of samples used in training. (C) OC-WCX2b, 95% of samples used in training. Increased probability classification thresholds increase Q5’s percent-correctly classified, positive predictive value, sensitivity, and specificity while decreasing the percent-classified.