Ranking Series of Cancer-Related Gene Expression Data by Means of the Superposing Significant Interaction Rules Method

Abstract

The Superposing Significant Interaction Rules (SSIR) method is a combinatorial procedure that deals with symbolic descriptors of samples. It is able to rank the series of samples when those items are classified into two classes. The method selects preferential descriptors and, with them, generates rules that make up the rank by means of a simple voting procedure. Here, two application examples are provided. In both cases, binary or multilevel strings encoding gene expressions are considered as descriptors. It is shown how the SSIR procedure is useful for ranking the series of patient transcription data to diagnose two types of cancer (leukemia and prostate cancer) obtaining Area Under Receiver Operating Characteristic (AU-ROC) values of 0.95 (leukemia prediction) and 0.80–0.90 (prostate). The preferential selected descriptors here are specific gene expressions, and this is potentially useful to point to possible key genes.

Application of SSIR Method for the Design of Fungicides

“…SSIR is variable selector [1,3] and finds a wide field of application in QSAR studies when dealing with families of congeneric compounds. Here, a set of congeneric fungicides [4] is analyzed.…”

“…The SSIR procedure has proven to be a useful tool in several fields. For instance, in chromatography variable selection [1], when modeling Phospholipidosis by using binary fingerprints [2] or in the treatment of transcriptomics related to leukemia and prostate cancer [3], among others.…”

“…Values of interest have been defined here as those of at least 80% percent of susceptibility (for a dose of 100 mg/L) 2. Values of interest: those with a percentage susceptibility of at least 100% (for a dose of 2.8 mg/L) 3. Values of interest: those of at least 80% susceptibility percentage (for a dose of 25 mg/L).…”