A method of classifying a test biological sample, preserved in a fixed state, from a human as to a cancer type using a support vector machine based classifier, said method comprising, at a computer system having at least one processor and memory storing at least one program for execution by the at least one processor to perform the method:
- (A) obtaining, for each respective gene xi in a first plurality of genes X={x1, . . . , xN}, a corresponding plurality Yi={yi1, . . . , yiM} of pairs of expression values for gene xi, each respective pair of expression values yij={yija, yijb} in Yi consisting of(i) a first expression value yija for the respective gene xi from a first human sample zia, preserved in a frozen state, and (ii) a second expression value yijb for the respective gene xi from a second human sample zib, preserved in a fixed state, in a first plurality of pairs of human samples Z={z1, . . . , zO}, wherein each sample in each respective pair of human samples zi={zia, zib} is from the same tissue type and the same subject as the other sample in the respective pair of human samples, and wherein N, M, and O are each positive integers and wherein M is 5 or greater;
- (B) determining for each respective pair of expression values yij={yija, yijb}, a corresponding plurality of correlation coefficients Wi={wi1, . . . , wiM} by comparing the first expression value yija for the respective gene xi from the first human sample zia, preserved in a frozen state, to the second expression value yijb for the respective gene xi from the second human sample zib, preserved in a fixed state;
- (C) selecting a second plurality of genes consisting of a sub-plurality of genes in the first plurality of genes using the criterion that each respective gene xi in the second plurality of genes has a correlation coefficient wij of 0.5 or higher for the first and second expression values {yija, yijb} across the plurality Yi of pairs of expression values for gene xi;
- (D) obtaining gene expression data from a second plurality of human samples, preserved in frozen states, each respective sample in the second plurality of human samples characterized by a cancer type selected from the group consisting of bladder, breast, colorectral, gastric, testicular germ cell, kidney, heptatocellular, non-small cell lung, non-Hodgkin's lymphoma, melanoma, ovarian, pancreatic, prostate, soft tissue sarcoma, and thyroid, wherein the gene expression data is stored in the memory;
- (E) training a support vector machine based classifier using (i) the gene expression data from the obtaining step (C), limited to a third plurality of genes, wherein at least ninety percent of the genes in the third plurality of genes are in the second plurality of genes, and (ii) the cancer type for each respective sample in the second plurality of samples, thereby constructing the support vector machine based classifier configured to determine the cancer type of the test biological human sample;
- (F) receiving abundance levels of respective transcripts of a group of genes in a nucleic acid preparation derived from said test biological sample;
- and
- (G) classifying said test biological sample as to said cancer type by applying said support vector machine based classifier constructed in the training step (D) to the abundance levels of the respective transcripts of the group of genes in the nucleic acid preparation derived from said test biological sample,
- wherein the obtaining step (A), the determining step (B), the selecting step (C), the obtaining step (D), the training step (E) and the processing step (F) are each performed using one or more suitably programmed computers.