|
|
|
|
How to Interpret the results-Making Your Choice
|
|
How to interpret the results
The result page contains three important values for evaluating each pTAA -TRP, OP and TSP. TRP defines the degree of restrictive expression of a given gene across human normal tissues and its confidence. we use known cancer testis antigens as training set to generate a detection call matrix, then tissue restriction score (TRS) was computed for each probe. The tissue restriction threshold was determined according to the 90% CT antigens' interval (Figure 1). Then the TRS of all the normal tissue datasets were gathered together and sorted by Unigene id. The tissue restriction penalty (TRP) was computed according to all the probes' TRS belong to each Unigene and their confidence evidenced by source sequences of probes and the amount of samples in the corresponding dataset. The higher the TRP, the more restrictive expression of a given gene across normal tissues.
Figure1: Percent of CT antigens (Y axis) with |TRS| lower than certain level (Show in X axis). Dotted line shows the cutoff for tissue restrictive genes with 90% CT antigens' interval.
Overexpression Penalty (OP) defines whether the expression of a given gene is unregulated significantly in cancerous tissues compared with corresponding normal tissues. Firstly, Differential expression analysis and significance test were done for each cancer microarray datasets and each cancer type of EST and SAGE expression data individually. The cancer/normal ratio was computed for each probe significantly differentially expressed. All of the cancer/normal ratio were gathered together and sorted by Unigene ID. Then OP was computed according to the overexpression ratio and their confidence evidenced by source sequences of probes. Thus the value of OP does not mean the differential expression ratio, but rather the combined view of differential expression ratio and the clues supporting the fact of overexpression. The higher the OP value, the more possibility of overexpression.
Tumor Specificity Penalty (TSP) gives an overall view of the tumor specificity. TSP was computed as following: TSP=TRP*OP. This algorism is designed according to the supposition that the tumor specificity increases in proportion with OP, and TRP. The higher the TSP, the higher degree of tumor specific expression.
When users are trying to find highly tumor specific genes, the three values should be considered together. The TSP gives an overall view, while the OP or TRS describes the overexpression and tissue restriction level respectively.
Users may found that each cancer type has a different OP and TSP. The reason why we drop a value for each cancer type was that many researchers actually want tumor specific genes overexpressed in their favored cancer type. Thus cancer type specific value may meet this purpose.
About Tissue Restriction
Users should be careful in interpreting the tissue restriction. As the normal tissue datasets cover most but not all human normal tissues, genes expressed restrictively in those uncovered normal tissues may be found with high TRP in the database. For example, UCLA normal tissue dataset do not contain leukocyte, genes specific to leukocyte may be found to be tissue restrictive in this case.
Users may found that some genes with high TRP belong to immune system specific genes. Irrespective of the problem of the tissue coverage of normal tissue datasets, we suspect that the upregulation of these genes may be originated from the tumor infiltration activity of immune cells. Anyway, there may be other unrecognized mechanism of the reason why these genes score high. Some researchers found that many tumor cells overexpress genes encoding antibodies with unknown specificity.
Another important notion was that tissue restrictive tumor antigens are all tumor specific genes, but not all tumor specific genes are tumor antigens. The preclude for a gene to be TAA was that the gene should be protein-coding. We give CCDS annotation for each gene to show which gene is protein-coding. But as more and more CD8+ T cell epitopes were discovered, it was found that some pseudogenes with multiple small ORFs can also be translated and provides a resource for T cell epitopes. Furthermore, Unigene clusters with no CCDS annotation should not be interpret as non-coding, as novel genes hiding in the current unigene assembly may not be contained in the CCDS database.
About the Immunogenicity-What Genes May Have Higher Possibility to be Immunogenic
Immunogenicity describes the ability of large molecules to elicit humeral or cellular immunity. The core of immunogenicity was heterogeneity, which means that an immunogen should not be recognized as self-antigen during the negative selection of the T cell repertoir. Tumor specificity actually reflect the heterogeneity with the preclude that the gene do encoding a protein and the protein expression correlates with mRNA expression. That is why we choose tumor specificity penalty as the core of the results.
Different branches of the immune system have different requirements for the protein product to elicit immunity. The humeral branch favors stable protein product larger than 10kD, while the cellular branch favors proteins with antigenic peptide, irrespective of the molecular weight of the protein product and its stability in the cell. Thus in order to be a B cell antigen, a gene must encode stable protein larger than 10kD; in order to be T cell antigen, a gene must encode T cell epitope containing protein or peptide.
From the view of anti-tumor effect of the immune response, only B cell antigen with membrane location may produce protective effect, while T cell antigen have no requirement on the subcellular location of the gene product in order to generate protective immunity.
In short, to select genes have the potential for the treatment of human carcinoma, genes with higher tumor specificity and higher tissue restriction level, membrane protein larger than 10kD or protein product with multiple predicted T cell epitope should be favored.
About the accuracy of this database
In our experience, the accuracy of this database should be about 10-20% in compared with RT-PCR, higher accuracy may be found when compared with northern blot or realtime PCR. Several factors contributes to this. First, the designation of a given sequence to a gene was done by EST clustering method, thus the correctness of the designation was determined by the accuracy of EST clustering. Secondly, different platforms have different deficiency in measuring the expression level. Thirdly, RT-PCR is generally more sensitive than any other platforms, thus tissue restrictive genes in the database may be found less restrictive after 35cycles of PCR amplification. As a result, we recommend realtime PCR instead of end point PCR in evaluating the tumor specificity of pTAA.
Functional Consideration
As more and more tumor associated antigens were found to be related to carcinogenesis, the functional aspects of tumor antigens gradually arise the immunologists' attention. As many organ-specific genes are found to be related to the function of their specifically expressed organs, together with the fact that tissue restrictive TAA are virtually tumor specific genes, it is not surprising to find that these genes contribute to the proliferation or metastasis of human carcinomas. Users can find many genes known to be related to carcinogenesis in our database. In order to help with users interested in functional aspects of cancer specific genes, we give an annotation of gene ontology for each gene in our database in the detail pages.
|
|
|
Making Choice
