Gene expression signature for classification of tissue of origin of tumor samples
A nucleic acid and sequence technology, applied in the field of cancer classification and determination of tissue origin, can solve problems that have not yet been determined
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment approach
[0087]According to one embodiment, the assay is based on the expression levels of 48 microRNAs in RNA extracted from FFPE metastatic tumor tissues. The test uses quantitative real-time reverse transcription polymerase chain reaction (qRT-PCR). RNA is first polyadenylated and then reverse-transcribed using a universal poly-T adapter to generate cDNA. The cDNA was detected with a specific forward primer and a universal reverse primer (complementary to the 5' end sequence of the poly T adapter) with a specific MGB probe (see the specific sequence in Table 1).
[0088] Analytical techniques for inferring sample origin by expression level include, but are not limited to: decision tree classifiers, logistic regression classifiers, linear regression classifiers, nearest neighbor classifiers (including K-nearest neighbors), neural network classifiers, and near-centroid classifiers.
[0089] Expression levels make binary decisions (at each relevant node) according to a pre-defined bin...
Embodiment 1
[0270] Samples and Spectrum
[0271] The discovery process profiles hundreds of samples to identify candidate biomarkers on an array platform. A training set of about 400 FFPE samples is used. RNA was extracted from these samples and subjected to qRT-PCR. Assays (Table 3, Figures 1-7) were constructed with 48 microRNAs to distinguish 26 classes representing 18 tissues of origin. An alternative assay was constructed, without identifying the bladder as an origin, ie, to distinguish 25 classes representing 17 tissues of origin.
[0272] A validation set of 255 new FFPE samples was used to evaluate test performance, representing 26 different tumor origins or "classes" (see Table 2 for a summary of the samples). In this set, about half of the samples were tumors that had metastasized to different sites (eg, lung, bone, brain, and liver). For all samples in the set, the proportion of tumors was at least 50%.
[0273] Table 2: Cancer Type, Class and Histology
[0274]
[027...
Embodiment 2
[0277] Decision Tree Classification Algorithm
[0278] Using the microRNA expression levels, a binary tree classification scheme (Fig. 1) was applied to construct a tumor classifier. This framework was built to exploit the specificity of microRNAs in tissue differentiation and embryonic development: different microRNAs are involved in different stages of tissue processing and are used by algorithms at different decision points or "nodes". The tree decomposes a complex multi-organizational classification problem into a series of simple binary decisions. At each node, no need to consider which type of branch sticks out of the tree earlier, reducing the interference of irrelevant samples and further simplifying the decision-making. Decision making at each node can then be done using only a small number of microRNA biomarkers with well-defined roles (see Table 3). The structure of the binary tree is based on the hierarchical and morphological similarity of tissue development 18...
PUM
| Property | Measurement | Unit |
|---|---|---|
| Sensitivity | aaaaa | aaaaa |
| Sensitivity | aaaaa | aaaaa |
Abstract
Description
Claims
Application Information
Login to View More