Given t 1 the number of information rich factors appears to be 4. Therefore, FA was performed with a growing number of such factors, from the one with higher selleck kinase inhibitor variance, up to 5, to test the appropriateness of the variance threshold. We then confirmed the validity of a subset of the Mod els using LDA to identify which factor was able to best classify tumor grade and histopathology, based on the statistical significance of Fisher exact test. This test, suited for contin gency tables where one or more expected frequencies are below 5, evaluates the null hypothesis associated with LDA that there are no statistically significant differ ences between the a priori clinically defined groups. The models for which the null hypothesis was rejected were retained. There fore, we performed 4 LDA, namely between a class and its complement, i.
e. high low grade, anaplastic non ana plastic, glioblastoma non glioblastoma and gliosarcoma non gliosarcoma, following the original classification in. We did not consider oligodendroglioma relevant, because of a single sample available. Model 3 appears to be the most suitable, since it is able to discriminate between anaplastic and non anaplastic tumors with 100% accuracy and the other two types of tumors with ? 92% accuracy. Since ana plastic tumors are low grade tumors, Factor 2 is relevant in the identification of low grade tumors in general with ? 92% accuracy, since the only oligodendroglioma appears to be elusive. It is worth noting that Model 4 shows the same performance scores, but with a greater number of factors and Factor 4 does not appear to be involved in class identification.
Interpretation of Multilevel Latent Structures mRNA Functional Analysis Working solely on Model 3, the mRNAs in each factor were processed to detect enriched Gene Ontology terms or UniProt keywords. The magni tude and sign of the factor scores give their relative relationship with the expression of miRNA and mRNA. Consequently, each row in the 3 factors score matrix was split into positive and negative portions and analyzed separately. F1 is associated with GO terms related to response to stress and external stimuli. Terms from SP keywords like secreted and glycoprotein were also found in this subset. Thus this factor appears then to be related with cell functions that process signal from the external environment to the cell with membrane receptors involved to the signal transduction.
F2? is also involved in the signaling, including categories related to cell adhe sion, it appears then to be related to functions like che motaxis that are involved in inflammation processes. Finally, F3 Cilengitide contains coding genes that are related to the biological process that goes under the general term gene expression. Gene expression includes all the mechanisms such as transcription, translation, RNA maturation, pro teins transport and ubiquitination by which information coded in the DNA is converted to a functional product.