Figure 1.
Classification across diseases.
The heatmap presented here distinguishes 4 different diseases using 70 peptides identified as the most significant using a 1-way ANOVA across the 27 breast cancer patients, 19 healthy controls, 10 Gliobastoma multiformae patients, and 9 Valley Fever patients. Classification accuracy was 100% using both linear discriminant analysis and Support Vector Machines and leave one out cross-validation.
Figure 2.
Glioblastoma training and test data.
The heatmap on the left shows 50 peptides that differentiated glioblastoma patients from healthy persons obtained from 4 different geographical locations across the US (Fred Hutchison Institute, University of Washington, University of California Irvine, Arizona State University). These peptides were also used to classify different samples consisting of blinded patient and healthy sera obtained from the Barrow Neurological Institute 3 years later. The colored bars on the right indicate clusters that define groups of peptides. Although there are differences between the values obtained in 2007 and 2010, most of the high-binding peptides are very similar.
Table 1.
Patient information and classification performance.
Figure 3.
Classification of multiple cancer types and molecular markers.
Top: six different classes of brain tumor patients were tested for their immunosignature. We examined Glioblastoma multiformae (MGMT- is brown, MGMT+ is purple), astrocytoma grade II (red), oligodendroglioma (cyan) and mixed oligo/astro (blue) against otherwise healthy controls (yellow). We used a 1-way ANOVA to select the 100 most significant peptides, p<10−18. High (red) and low (blue) signals correspond to patient antibodies detected with a fluorescently labeled anti-human secondary. Data was grouped using hierarchical clustering on both peptides (Y-axis) and patients (X-axis). Bottom right: principal components display of the separation between samples. X and Y axes represent the first two principal components making up 64% of the total variance across the samples. Patient information is found in Table 1.