False Discovery Rate Calculation Research Articles

Absence of MxA induction by interferon β in patients with MS reflects complete loss of bioactivity

In patients with multiple sclerosis (MS), neutralizing antibodies (NAbs) appearing during treatment with interferon (IFN) beta reduce or in high concentrations abolish bioactivity and therapeutic efficacy. In vivo MxA induction by IFNbeta is used as a marker of biologic response to IFNbeta. It has been argued that despite absence of MxA induction measured by PCR, some bioactivity might be preserved. In a cohort study, we measured gene expression by gene chip analysis in NAb-negative and NAb-positive patients to test that hypothesis. The effect of IFNbeta was studied by comparing samples collected before and 9-12 hours after an injection. The cohort consisted of 12 NAb-positive patients without MxA response and 12 NAb-negative patients with preserved response. MxA in vivo response was determined in whole blood using real-time PCR. Screening for IFNbeta-regulated genes in mononuclear cells was done using gene chips. False discovery rate (FDR) analysis was used as statistical tool. Of 8,793 genes, 5,593 were detectable in at least one patient in both groups. Of these, calculation of FDR revealed 1,077 IFNbeta-regulated genes at a 5% level in NAb-negative patients. The corresponding number of IFNbeta-regulated genes in NAb-positive patients was zero. In neutralizing antibody (NAb)-positive patients without an MxA response, we were not able to detect differential expression of any of the 1077 interferon (IFN) beta-regulated genes identified in NAb-negative patients. Lack of MxA in vivo response in patients with multiple sclerosis with NAbs is a reliable marker of a completely blocked biologic response to IFNbeta, with no indication of residual bioactivity.

An easy‐to‐use Decoy Database Builder software tool, implementing different decoy strategies for false discovery rate calculation in automated MS/MS protein identifications

One of the major challenges for large scale proteomics research is the quality evaluation of results. Protein identification from complex biological samples or experimental setups is often a manual and subjective task which lacks profound statistical evaluation. This is not feasible for high-throughput proteomic experiments which result in large datasets of thousands of peptides and proteins and their corresponding mass spectra. To improve the quality, reliability and comparability of scientific results, an estimation of the rate of erroneously identified proteins is advisable. Moreover, scientific journals increasingly stipulate that articles containing considerable MS data should be subject to stringent statistical evaluation. We present a newly developed easy-to-use software tool enabling quality evaluation by generating composite target-decoy databases usable with all relevant protein search engines. This tool, when used in conjunction with relevant statistical quality criteria, enables to reliably determine peptides and proteins of high quality, even for nonexperienced users (e.g. laboratory staff, researchers without programming knowledge). Different strategies for building decoy databases are implemented and the resulting databases are characterized and compared. The quality of protein identification in high-throughput proteomics is usually measured by the false positive rate (FPR), but it is shown that the false discovery rate (FDR) delivers a more meaningful, robust and comparable value.

Experimental and Statistical Considerations to Avoid False Conclusions in Proteomics Studies Using Differential In-gel Electrophoresis

In quantitative proteomics, the false discovery rate (FDR) can be defined as the number of false positives within statistically significant changes in expression. False positives accumulate during the simultaneous testing of expression changes across hundreds or thousands of protein or peptide species when univariate tests such as the Student's t test are used. Currently most researchers rely solely on the estimation of p values and a significance threshold, but this approach may result in false positives because it does not account for the multiple testing effect. For each species, a measure of significance in terms of the FDR can be calculated, producing individual q values. The q value maintains power by allowing the investigator to achieve an acceptable level of true or false positives within the calls of significance. The q value approach relies on the use of the correct statistical test for the experimental design. In this situation, a uniform p value frequency distribution when there are no differences in expression between two samples should be obtained. Here we report a bias in p value distribution in the case of a three-dye DIGE experiment where no changes in expression are occurring. The bias was shown to arise from correlation in the data from the use of a common internal standard. With a two-dye schema, where each sample has its own internal standard, such bias was removed, enabling the application of the q value to two different proteomics studies. In the case of the first study, we demonstrate that 80% of calls of significance by the more traditional method are false positives. In the second, we show that calculating the q value gives the user control over the FDR. These studies demonstrate the power and ease of use of the q value in correcting for multiple testing. This work also highlights the need for robust experimental design that includes the appropriate application of statistical procedures.

Large-Scale Evaluation of Candidate Genes Identifies Associations between VEGF Polymorphisms and Bladder Cancer Risk

Common genetic variation could alter the risk for developing bladder cancer. We conducted a large-scale evaluation of single nucleotide polymorphisms (SNPs) in candidate genes for cancer to identify common variants that influence bladder cancer risk. An Illumina GoldenGate assay was used to genotype 1,433 SNPs within or near 386 genes in 1,086 cases and 1,033 controls in Spain. The most significant finding was in the 5′ UTR of VEGF (rs25648, p for likelihood ratio test, 2 degrees of freedom = 1 × 10−5). To further investigate the region, we analyzed 29 additional SNPs in VEGF, selected to saturate the promoter and 5′ UTR and to tag common genetic variation in this gene. Three additional SNPs in the promoter region (rs833052, rs1109324, and rs1547651) were associated with increased risk for bladder cancer: odds ratio (95% confidence interval): 2.52 (1.06–5.97), 2.74 (1.26–5.98), and 3.02 (1.36–6.63), respectively; and a polymorphism in intron 2 (rs3024994) was associated with reduced risk: 0.65 (0.46–0.91). Two of the promoter SNPs and the intron 2 SNP showed linkage disequilibrium with rs25648. Haplotype analyses revealed three blocks of linkage disequilibrium with significant associations for two blocks including the promoter and 5′ UTR (global p = 0.02 and 0.009, respectively). These findings are biologically plausible since VEGF is critical in angiogenesis, which is important for tumor growth, its elevated expression in bladder tumors correlates with tumor progression, and specific 5′ UTR haplotypes have been shown to influence promoter activity. Associations between bladder cancer risk and other genes in this report were not robust based on false discovery rate calculations. In conclusion, this large-scale evaluation of candidate cancer genes has identified common genetic variants in the regulatory regions of VEGF that could be associated with bladder cancer risk.

Model-based analysis of tiling-arrays for ChIP-chip

We propose a fast and powerful analysis algorithm, titled Model-based Analysis of Tiling-arrays (MAT), to reliably detect regions enriched by transcription factor chromatin immunoprecipitation (ChIP) on Affymetrix tiling arrays (ChIP-chip). MAT models the baseline probe behavior by considering probe sequence and copy number on each array. It standardizes the probe value through the probe model, eliminating the need for sample normalization. MAT uses an innovative function to score regions for ChIP enrichment, which allows robust P value and false discovery rate calculations. MAT can detect ChIP regions from a single ChIP sample, multiple ChIP samples, or multiple ChIP samples with controls with increasing accuracy. The single-array ChIP region detection feature minimizes the time and monetary costs for laboratories newly adopting ChIP-chip to test their protocols and antibodies and allows established ChIP-chip laboratories to identify samples with questionable quality that might contaminate their data. MAT is developed in open-source Python and is available at http://chip.dfci.harvard.edu/ approximately wli/MAT. The general framework presented here can be extended to other oligonucleotide microarrays and tiling array platforms.

Relationship between tumor gene expression and recurrence in stage II/III colon cancer: Quantitative RT-PCR assay of 757 genes in fixed paraffin-embedded (FPE) tissue

3518 Background: As an initial step in developing abetter method of assessing prognosis following potentially curative surgery for colon cancer, we performed an exploratory gene identification study. Methods: RNA was extracted from three 10-micron sections of the FPE tumor tissue obtained at the initial diagnosis from 353 patients entered between 1977 and 1984 into the surgery-only or surgery-plus BCG arms of NSABP C-01/C-02. Expression was quantified for 757 cancer-related and reference genes with RT-PCR. Results: Blocks from 270 patients were evaluable after pre-specified exclusions: 128 were stage II and 142 were stage III. All patients had ≥ 5-year follow up. In univariate Cox proportional hazard analyses, 148 genes exhibited a nominally significant (unadjusted p-value <0.05) linear association with recurrence-free interval (RFI) (7 genes p≤0.001, 66 genes 0.001<p<0.01; 75 genes 0.01≤p<0.05). False discovery rate calculations suggest that about 25% of the 148 genes are expected to be false positives. Higher expression was associated with shorter RFI for 118 genes including SERPINB5, DUSP1, AKT3, TIMP1, ANXA2, RHOB (p≤ 0.001) and with longer RFI for 30 genes including BRCA1 (p≤0.001). The relationship between gene expression and RFI was similar for stage II and stage III patients for 143 of the 148 genes found to be significant. The largest cluster of genes is functionally related to extracellular matrix remodeling. The magnitude of the hazard ratios is similar to that observed in the early Oncotype DX studies in breast cancer and should allow clinically useful separation into low/intermediate/high risk groups. Conclusions: Quantitative RT-PCR assay of FPE colon cancer tissue can be used to identify large numbers of genes associated with RFI in patients with stage II and III colon cancer. If these results are confirmed by additional studies in progress, this technique has promise to improve selection of colon cancer patients for adjuvant chemotherapy. Funded in part as a collaborative arrangement under contract between the NSABP Foundation, Inc., and Genomic Health, Inc. [Table: see text]

Identification of gene expression profiles predicting tumor cell response to l-alanosine

The methylthioadenosine phosphorylase (MTAP) gene gained considerable interest as therapeutic target for tumors with the 9p21 deletion. This gene maps to 9p21 and loss of this chromosomal region in tumors offers an unique opportunity for chemoselective treatment, since MTAP is an important salvage enzyme for the formation of adenine that is needed for DNA synthesis. l-Alanosine, an antibiotic from Streptomyces alanosinicus, blocks the common de novo purine biosynthesis pathway and, thereby, inhibits tumor cells with MTAP deficiency. Normal cells escape the detrimental effects of l-alanosine due to their proficiency in the MTAP salvage pathway. The present analysis was undertaken to gain insights into the molecular architecture of tumor cells that determines the response to l-alanosine apart from the MTAP gene. Analysis of cell doubling times and ic50 values for l-alanosine showed that slowly growing cell lines were more resistant to l-alanosine than rapidly growing ones. Mining the database of the National Cancer Institute (N.C.I.), for the mRNA expression of 9706 genes in 60 cell lines by means of Kendall’s τ-test, false discovery rate calculation, and hierarchical cluster analysis pointed to 11 genes or expressed sequence tags whose mRNA expression correlated with the ic50 values for l-alanosine. Furthermore, we tested l-alanosine for cross-resistance in multidrug-resistant cell lines which overexpress selectively either the P-glycoprotein/MDR1 (CEM/ADR5000), MRP1 (HL-60/AR), or BCRP (MDA-MB-231-BCRP) genes. None of the multidrug-resistant cell lines was cross-resistant to l-alanosine indicating that l-alanosine may be suitable to treat multidrug-resistant, refractory tumors in the clinic. Finally, the ic50 values for l-alanosine of the 60 cell lines were correlated to the p53 mutational status and expression of p53 downstream genes. We found that p53 mutated cell lines were more resistant to l-alanosine than p53 wild type cell lines.