AIM:To investigate the diverse characteristics of different pathological gradings of gastric adenocarcinoma (GA) using tumor-related genes.METHODS:GA tissues in different pathological gradings and normal tissues were subjected to tissue arrays.Expressions of 15 major tumor-related genes were detected by RNA in situ hybridization along with 3’ terminal digoxin-labeled anti-sense single strandedoligonucleotide and locked nucleic acid modifying probe within the tissue array.The data obtained were processed by support vector machines by four different feature selection methods to discover the respective critical gene/gene subsets contributing to the GA activities of different pathological gradings.RESULTS:In comparison of poorly differentiated GA with normal tissues,tumor-related gene TP53 plays a key role,although other six tumor-related genes could also achieve the Area Under Curve (AUC) of the receiver operating characteristic independently by more than 80%.Comparing the well differentiated GA with normal tissues,we found that 11 tumor-related genes could independently obtain the AUC by more than 80%,but only the gene subsets,TP53,RB and PTEN,play a key role.Only the gene subsets,Bcl10,UVRAG,APC,Beclin1,NM23,PTEN and RB could distinguish between the poorly differentiated and well differentiated GA.None of a single gene could obtain a valid distinction.CONCLUSION:Different from the traditional point of view,the well differentiated cancer tissues have more alterations of important tumor-related genes than the poorly differentiated cancer tissues. To investigate the diverse characteristics of different pathological gradings of gastric adenocarcinoma (GA) using tumor-related genes. METHODS: GA tissues in different pathological gradings and normal tissues were subjected to tissue arrays. Expressions of 15 major tumor-related genes were detected by RNA in situ hybridization along with 3 ’terminal digoxin-labeled anti-sense single stranded oligonucleotide and locked nucleic acid modifying probe within the tissue array. The data obtained were processed by support vector machines by four different feature selection methods to discover the particular critical gene / gene subsets contributing to the GA activities of different pathological gradings .RESULTS: In comparisonCompare poorly differentiated GA with normal tissues, tumor-related gene TP53 plays a key role, although other six tumor-related genes could also achieve the Area Under Curve ( AUC) of the receiver operating characteristic independently independently by more than 80% ated GA with normal tissues, we found that 11 tumor-related genes could have obtained the AUC by more than 80%, but only the gene subsets, TP53, RB and PTEN, play a key role. Notly the gene subsets, Bcl10, UVRAG , APC, Beclin1, NM23, PTEN and RB could distinguish between the poorly differentiated and well differentiated GA. Single of a single gene could have a valid distinction. CONCLUSION: Different from the traditional point of view, the well differentiated cancer tissues have more alterations of important tumor-related genes than the poorly differentiated cancer tissues.