Dataset Information

Classifying gene expression profiles from pairwise mRNA comparisons.

ABSTRACT: We present a new approach to molecular classification based on mRNA comparisons. Our method, referred to as the top-scoring pair(s) (TSP) classifier, is motivated by current technical and practical limitations in using gene expression microarray data for class prediction, for example to detect disease, identify tumors or predict treatment response. Accurate statistical inference from such data is difficult due to the small number of observations, typically tens, relative to the large number of genes, typically thousands. Moreover, conventional methods from machine learning lead to decisions which are usually very difficult to interpret in simple or biologically meaningful terms. In contrast, the TSP classifier provides decision rules which i) involve very few genes and only relative expression values (e.g., comparing the mRNA counts within a single pair of genes); ii) are both accurate and transparent; and iii) provide specific hypotheses for follow-up studies. In particular, the TSP classifier achieves prediction rates with standard cancer data that are as high as those of previous studies which use considerably more genes and complex procedures. Finally, the TSP classifier is parameter-free, thus avoiding the type of over-fitting and inflated estimates of performance that result when all aspects of learning a predictor are not properly cross-validated.

SUBMITTER: Geman D

PROVIDER: S-EPMC1989150 | biostudies-literature | 2004

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Classifying gene expression profiles from pairwise mRNA comparisons.

Geman Donald D d'Avignon Christian C Naiman Daniel Q DQ Winslow Raimond L RL

Statistical applications in genetics and molecular biology 20040830

We present a new approach to molecular classification based on mRNA comparisons. Our method, referred to as the top-scoring pair(s) (TSP) classifier, is motivated by current technical and practical limitations in using gene expression microarray data for class prediction, for example to detect disease, identify tumors or predict treatment response. Accurate statistical inference from such data is difficult due to the small number of observations, typically tens, relative to the large number of g ...[more]

PMID: 16646797

Similar Datasets

Project description:BackgroundGene expression microarray technology permits the analysis of global gene expression profiles. The amount of sample needed limits the use of small excision biopsies and/or needle biopsies from human or animal tissues. Linear amplification techniques have been developed to increase the amount of sample derived cDNA. These amplified samples can be hybridised on microarrays. However, little information is available whether microarrays based on amplified and unamplified material yield comparable results. In the present study we compared microarray data obtained from amplified mRNA derived from biopsies of rat cardiac left ventricle and non-amplified mRNA derived from the same organ. Biopsies were linearly amplified to acquire enough material for a microarray experiment. Both amplified and unamplified samples were hybridized to the Rat Expression Set 230 Array of Affymetrix.ResultsAnalysis of the microarray data showed that unamplified material of two different left ventricles had 99.6% identical gene expression. Gene expression patterns of two biopsies obtained from the same parental organ were 96.3% identical. Similarly, gene expression pattern of two biopsies from dissimilar organs were 92.8% identical to each other.Twenty-one percent of reporters called present in parental left ventricular tissue disappeared after amplification in the biopsies. Those reporters were predominantly seen in the low intensity range. Sequence analysis showed that reporters that disappeared after amplification had a GC-content of 53.7+/-4.0%, while reporters called present in biopsy- and whole LV-samples had an average GC content of 47.8+/-5.5% (P <0.001). Those reporters were also predicted to form significantly more (0.76+/-0.07 versus 0.38+/-0.1) and longer (9.4+/-0.3 versus 8.4+/-0.4) hairpins as compared to representative control reporters present before and after amplification.ConclusionThis study establishes that the gene expression profile obtained after amplification of mRNA of left ventricular biopsies is representative for the whole left ventricle of the rat heart. However, specific gene transcripts present in parental tissues were undetectable in the minute left ventricular biopsies. Transcripts that were lost due to the amplification process were not randomly distributed, but had higher GC-content and hairpins in the sequence and were mainly found in the lower intensity range which includes many transcription factors from specific signalling pathways.

Project description:BackgroundValidation of previously identified candidate biomarkers and identification of additional candidate gene expression profiles to facilitate diagnosis of tuberculosis (TB) disease and monitoring treatment responses in the Ethiopian context is vital for improving TB control in the future.MethodsExpression levels of 105 immune-related genes were determined in the blood of 80 HIV-negative study participants composed of 40 active TB cases, 20 latent TB infected individuals with positive tuberculin skin test (TST+), and 20 healthy controls with no Mycobacterium tuberculosis (Mtb) infection (TST-), using focused gene expression profiling by dual-color Reverse-Transcription Multiplex Ligation-dependent Probe Amplification assay. Gene expression levels were also measured six months after anti-TB treatment (ATT) and follow-up in 38 TB patients.ResultsThe expression of 15 host genes in TB patients could accurately discriminate between TB cases versus both TST+ and TST- controls at baseline and thus holds promise as biomarker signature to classify active TB disease versus latent TB infection in an Ethiopian setting. Interestingly, the expression levels of most genes that markedly discriminated between TB cases versus TST+ or TST- controls did not normalize following completion of ATT therapy at 6 months (except for PTPRCv1, FCGR1A, GZMB, CASP8 and GNLY) but had only fully normalized at the 18 months follow-up time point. Of note, network analysis comparing TB-associated host genes identified in the current HIV-negative TB cohort to TB-associated genes identified in our previously published Ethiopian HIV-positive TB cohort, revealed an over-representation of pattern recognition receptors including TLR2 and TLR4 in the HIV-positive cohort which was not seen in the HIV-negative cohort. Moreover, using ROC cutoff ≥ 0.80, FCGR1A was the only marker with classifying potential between TB infection and TB disease regardless of HIV status.ConclusionsOur data indicate that complex gene expression signatures are required to measure blood transcriptomic responses during and after successful ATT to fully diagnose TB disease and characterise drug-induced relapse-free cure, combining genes which resolve completely during the 6-months treatment phase of therapy with genes that only fully return to normal levels during the post-treatment resolution phase.

Dataset Information

Classifying gene expression profiles from pairwise mRNA comparisons.

Publications

Classifying gene expression profiles from pairwise mRNA comparisons.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets