IADR Abstract Archives
Are There Gastric Cancer Biomarkers in Salivary Transcriptome?
Objectives: By utilizing the salivary transcriptomic data, discovery of biomarkers that could aid in early detection of gastric cancer (GC) through saliva sampling is feasible. This current study has analyzed salivary transcriptome data with a view to identify potential GC biomarkers.
Methods: Extensive analyzation of microarray data sets has been conducted in order to understand the complexity and heterogeneity associated with GC. To identify features that allow discernment of GC and non-GC cases, a statistical multiple comparisons-based approach has been used. These were then used to build an artificial neural network-based machine learning model. Additionally, the machine learning model was used to delineate features critical for its internal representation by extending a calliper randomization approach to the analysis of microarray data. Functional analysis of the features identified by the calliper randomization approach revealed genes that have been associated with GC.
Results: Detailed comparison of the transcript level quantified in gastric cancer samples and normal samples revealed that 2,257 features were differentially expressed between the two classes of samples. There were 241 features that were significantly different within gastric cancer samples and 316 features that were significantly different within normal samples. These results do not point to a significant variability within the samples.The analysis of the features for molecular components revealed that the features belonged to four categories viz. cadherin binding, enzyme inhibitory activity, magnesium ion binding and ubiquitin-protein transferase activity.
Conclusions: Analysis of salivary transcriptome data using statistical and machine learning-based methods revealed features capable of distinguishing GC samples from normal samples. This did not reveal significant heterogeneity within the samples. Using the ANN-based calliper randomization approach, it was possible to identify features capable of classifying the samples perfectly. Functional analysis of the data revealed that the genes associated with the features were associated with GC. These results point to the fact that there would be significant benefits in further investigating the use saliva as a source for non-invasive early detection of GC.
Methods: Extensive analyzation of microarray data sets has been conducted in order to understand the complexity and heterogeneity associated with GC. To identify features that allow discernment of GC and non-GC cases, a statistical multiple comparisons-based approach has been used. These were then used to build an artificial neural network-based machine learning model. Additionally, the machine learning model was used to delineate features critical for its internal representation by extending a calliper randomization approach to the analysis of microarray data. Functional analysis of the features identified by the calliper randomization approach revealed genes that have been associated with GC.
Results: Detailed comparison of the transcript level quantified in gastric cancer samples and normal samples revealed that 2,257 features were differentially expressed between the two classes of samples. There were 241 features that were significantly different within gastric cancer samples and 316 features that were significantly different within normal samples. These results do not point to a significant variability within the samples.The analysis of the features for molecular components revealed that the features belonged to four categories viz. cadherin binding, enzyme inhibitory activity, magnesium ion binding and ubiquitin-protein transferase activity.
Conclusions: Analysis of salivary transcriptome data using statistical and machine learning-based methods revealed features capable of distinguishing GC samples from normal samples. This did not reveal significant heterogeneity within the samples. Using the ANN-based calliper randomization approach, it was possible to identify features capable of classifying the samples perfectly. Functional analysis of the data revealed that the genes associated with the features were associated with GC. These results point to the fact that there would be significant benefits in further investigating the use saliva as a source for non-invasive early detection of GC.