MicroRNA-transcription factor interactions and their combined effect on target gene expression in colon cancer cases.

Transcription factors (TFs) and microRNAs (miRNAs) regulate gene expression: TFs by influencing messenger RNA (mRNA) transcription and miRNAs by influencing mRNA translation and transcript degradation. Additionally, miRNAs and TFs alter each other's expression, making it difficult to ascertain the effect either one has on target gene (TG) expression. In this investigation, we use a two-way interaction model with the TF and miRNA as independent variables to investigate whether miRNAs and TFs work together to influence TG expression levels in colon cancer subjects. We used known TF binding sites and validated miRNA targets to determine potential miRNA-TF-TG interactions, restricting interactions to those with a TF previously associated with altered risk of colorectal cancer death. We analyzed interactions using normal colonic mucosa expression as well as differential expression, which is measured as colonic carcinoma expression minus normal colonic mucosa expression. We analyzed 3518 miRNA-TF-TG triplets using normal mucosa expression and 617 triplets using differential expression. Normal colonic RNA-Seq data were available for 168 individuals; of these, 159 also had carcinoma RNA-Seq data. Thirteen unique miRNA-TF-TG interactions, comprising six miRNAs, four TFs, and 11 TGs, were statistically significant after adjustment for multiple comparisons in normal colonic mucosa, and 14 unique miRNA-TF-TG interactions, comprising two miRNAs, two TFs, and 13 TGs, were found for carcinoma-normal differential expression. Our results show that TG expression is influenced by both miRNAs as well as TFs, and the influence of one regulator impacts the effect of the other on the shared TG expression.

Power in pairs: assessing the statistical value of paired samples in tests for differential expression.

BACKGROUND: When genomics researchers design a high-throughput study to test for differential expression, some biological systems and research questions provide opportunities to use paired samples from subjects, and researchers can plan for a certain proportion of subjects to have paired samples. We consider the effect of this paired samples proportion on the statistical power of the study, using characteristics of both count (RNA-Seq) and continuous (microarray) expression data from a colorectal cancer study. RESULTS: We demonstrate that a higher proportion of subjects with paired samples yields higher statistical power, for various total numbers of samples, and for various strengths of subject-level confounding factors. In the design scenarios considered, the statistical power in a fully-paired design is substantially (and in many cases several times) greater than in an unpaired design. CONCLUSIONS: For the many biological systems and research questions where paired samples are feasible and relevant, substantial statistical power gains can be achieved at the study design stage when genomics researchers plan on using paired samples from the largest possible proportion of subjects. Any cost savings in a study design with unpaired samples are likely accompanied by underpowered and possibly biased results.

The PI3K/AKT signaling pathway: Associations of miRNAs with dysregulated gene expression in colorectal cancer.

The PI3K/AKT-signaling pathway is one of the most frequently activated signal-transduction pathways in cancer. We examined how dysregulated gene expression is associated with miRNA expression in this pathway in colorectal cancer (CRC). We used data from 217 CRC cases to evaluate differential pathway gene expression between paired carcinoma and normal mucosa and identify miRNAs that are associated with these genes. Gene expression data from RNA-Seq and miRNA expression data from Agilent Human miRNA Microarray V19.0 were analyzed. We focused on genes most associated with CRC (fold change (FC) of >1.5 or <0.67) that were statistically significant after adjustment for multiple comparisons. Of the 304 genes evaluated, 76 had a FC of <0.67, and 57 had a FC of >1.50; 47 of these genes were associated with miRNA differential expression. There were 145 mRNA:miRNA seed-region matches of which 26 were inversely associated suggesting a greater likelihood of a direct association. Most miRNA:mRNA associations were with factors that stimulated the pathway. For instance, both IL6R and PDGFRA had inverse seed-region matches with seven miRNAs, suggesting that these miRNAs have a direct effect on these genes and may be key elements in activation of the pathway. Other miRNA:mRNA associations with similar impact on the pathway were miR-203a with ITGA4, miR-6071 with ITGAV, and miR-375 with THBS2, all genes involved in extracellular matrix function that activate PI3Ks. Gene expression in the PI3K/Akt-signaling pathway is dysregulated in CRC. MiRNAs were associated with many of these dysregulated genes either directly or in an indirect manner.

Infrequently expressed miRNAs in colorectal cancer tissue and tumor molecular phenotype.

This corrects the article DOI: 10.1038/modpathol.2017.38.

The miRNA landscape of colorectal polyps.

The genomic landscape of adenomas and polyps may help define disease pathways. Expression of miRNAs in adenomas and polyps may importantly contribute to these pathways. We evaluated miRNA expression in 293 polyp-normal colorectal mucosa pairs. Polyps were classified as either adenomatous polyp (AD), hyperplastic polyp (HP), or sessile serrated polyp (SSP). We compared these miRNA expression profiles in polyps to miRNA expression in microsatellite unstable (MSI) and stable (MSS) tumors. A False Discovery Rate of 0.05 based on Benjamini and Hochberg was used to adjust for multiple comparisons. There were 70 miRNAs with differential expression by polyp type with a fold change <0.75 or >1.34 after adjusting for multiple comparisons. The major differences in miRNA expression were observed between AD and SSP and AD and HP, with few differences in expression noted for SSP and HP. AD polyps were more likely to be upregulated from normal colonic mucosa, while SSP and HP were more likely to be downregulated from normal colonic mucosa. MiRNA expression in the SSP and HP tumors almost uniformly go in opposite directions from the MSS tumor miRNA expression and was mixed with MSI tumors. We conclude that different types of polyps have unique miRNA expression profiles. © 2017 Wiley Periodicals, Inc.

Transcription factor-microRNA associations and their impact on colorectal cancer survival.

MicroRNAs (miRNAs) and Transcription Factors (TFs) both influence messenger RNA (mRNA) expression, disrupting biological pathways involved in carcinogenesis and prognosis. As many miRNAs target multiple mRNAs, thus influencing a multitude of biological pathways, deciphering which miRNAs are important for cancer development and survival is difficult. In this study, we (i) determine associations between TF and survival (N = 168 colon cancer cases); (ii) identify miRNAs associated with TFs related to survival; and (iii) determine if factors derived from TF-specific miRNA principal component analysis (PCA) influence survival. Cox Proportional hazard models were run for each PCA factor to determine Hazard Ratios (HR) and 95% Confidence Intervals (CI) adjusting for age, center, and AJCC stage. Thirty TFs improved survival when differential expression increased; 27 of these were associated significantly with normal colonic mucosa expression of 65 unique miRNAs when an FDR q-value of <0.05 was applied. Five factors, comprising 21 miRNAs, altered survival in rectal cancer subjects; four of these five factors improved survival and one factor reduced survival. One factor comprising four miRNAs reduced survival in colon cancer subjects. In summary, our data suggest that expression of TFs and their related miRNAs influence survival after diagnosis with colorectal cancer.

Alterations in microRNA expression associated with alcohol consumption in rectal cancer subjects.

PURPOSE: Alcohol consumption has been purported to influence many diseases. MicroRNAs (miRNAs) may be influenced by compounds found in alcohol. In this investigation, we test the hypothesis that total alcohol, beer, wine, and hard liquor influence miRNA expression. METHODS: We studied 1447 colorectal (CR) cancer cases with normal CR mucosa and carcinoma miRNA expression data along with alcohol consumption data. We analyzed long-term and long-term and current (LTC) alcohol use for beer, liquor, and wine with miRNA expression between paired carcinoma and normal colon and rectal tissues, adjusting for multiple comparisons using the positive false discovery rate q-value. MiRNAs associated significantly with alcohol were examined with all-cause mortality (ACM). MiRNAs associated significantly with ACM were examined with RNA-Seq data. RESULTS: Expression of 84 miRNAs was associated significantly with LTC wine use in normal rectal mucosa. Higher expression of two of these miRNAs significantly worsened ACM: hsa-miR-210 (Hazard Ratio [HR] 1.12, 95% CI (1.03, 1.21), p-value = 0.004), and hsa-miR-92a-1-5p (HR 1.20, 95% CI (1.04, 1.38), p-value = 0.013). These miRNAs were downregulated across levels of LTC wine consumption. CONCLUSIONS: Our results suggest that wine influences miRNA expression in rectal cancer, supporting the hypothesis that components in alcohol influence miRNA expression.

Infrequently expressed miRNAs in colorectal cancer tissue and tumor molecular phenotype.

We have previously shown that commonly expressed miRNAs influenced tumor molecular phenotype in colorectal cancer. We hypothesize that infrequently expressed miRNAs, when showing higher levels of expression, help to define tumor molecular phenotype. In this study, we examine 304 miRNAs expressed in at least 30 individuals, but in <50% of the population and with a mean level of expression above 1.0 relative florescent unit. We examine associations in 1893 individuals who have the tumor molecular phenotype data as well as miRNA expression levels for both carcinoma and normal colorectal tissue. We compare miRNAs uniquely associated with tumor molecular phenotype to the RNAseq data to identify genes associated with these miRNAs. This information is used to further identify unique pathways associated with tumor molecular phenotypes of TP53-mutated, KRAS-mutated, CpG island methylator phenotype and microsatellite instability tumors. Thirty-seven miRNAs were uniquely associated with TP53-mutated tumors; 30 of these miRNAs had higher level of expression in TP53-mutated tumors, while seven had lower levels of expression. Of the 34 miRNAs associated with CpG island methylator phenotype-high tumors, 16 were more likely to have a CpG island methylator phenotype-high tumor and 19 were less likely to be CpG island methylator phenotype-high. For microsatellite instability, 13 of the 22 infrequently expressed miRNAs were significantly less likely to be expressed in microsatellite unstable tumors. KRAS-mutated tumors were not associated with any miRNAs after adjustment for multiple comparisons. Of the dysregulated miRNAs, 17 were more likely to be TP53-mutated tumors while simultaneously being less likely to be CpG island methylator phenotype-high and/or microsatellite instability tumors. Genes regulated by these miRNAs were involved in numerous functions and pathways that influence cancer risk and progression. In summary, some infrequently expressed miRNAs, when expressed at higher levels, appear to have significant biological meaning in terms of tumor molecular phenotype and gene expression profiles.

Identifying factors associated with the direction and significance of microRNA tumor-normal expression differences in colorectal cancer.

BACKGROUND: microRNAs are small non-protein-coding RNA molecules that regulate gene expression, and have a potential epigenetic role in disease progression and survival of colorectal cancer. In terms of tumor-normal expression differences, many microRNAs exhibit evidence of being up-regulated in some subjects but down-regulated in others, or are dysregulated only for a subset of the population. We present and implement an approach to identify factors (lifestyle, tumor molecular phenotype, and survival-related) that are associated with the direction and/or significance of these microRNAs' tumor-normal expression differences in colorectal cancer. METHODS: Using expression data for 1394 microRNAs and 1836 colorectal cancer subjects (each with both tumor and normal samples), we perform a dip test to identify microRNAs with multimodal distributions of tumor-normal expression differences. For proximal, distal, and rectal tumor sites separately, these microRNAs are tested for tumor-normal differential expression using a signed rank test, both overall and within levels of each lifestyle, tumor molecular phenotype, and survival-related factor. Appropriate adjustments are made to control the overall FDR. RESULTS: We identify hundreds of microRNAs whose direction and/or significance of tumor-normal differential expression is associated with one or more lifestyle, tumor molecular phenotype, or survival-related factors. CONCLUSIONS: The results of this study demonstrate the benefit to colorectal cancer researchers to consider multiple subject-level factors when studying dysregulation of microRNAs, whose tumor-related changes in expression can be associated with multiple factors. Our results will serve as a publicly-available resource to provide clarifying information about various factors associated with the direction and significance of tumor-normal differential expression of microRNAs in colorectal cancer.

MicroRNA profiles in colorectal carcinomas, adenomas and normal colonic mucosa: variations in miRNA expression and disease progression.

MiRNAs are small, non-protein-coding RNA molecules that regulate gene expression either by post-transcriptionally suppressing mRNA translation or by mRNA degradation. We examine differentially expressed miRNAs in colorectal carcinomas, adenomas and normal colonic mucosa. Data come from population-based studies of colorectal cancer conducted in Utah and the Kaiser Permanente Medical Care Program. A total of 1893 carcinoma/normal-paired samples and 290 adenoma tissue samples were run on the Agilent Human miRNA Microarray V19.0 which contained 2006 miRNAs. We tested for significant differences in miRNA expression between paired carcinoma/adenoma/normal colonic tissue samples. Fewer than 600 miRNAs were expressed in >80% of people for colonic tissue; of these 86.5% were statistically differentially expressed between carcinoma and normal colonic mucosa using a false discovery rate of 0.05. Roughly half of these differentially expressed miRNAs showed a progression in levels of expression from normal to adenoma to carcinoma tissue. Other miRNAs appeared to be altered at the normal to adenoma stage, while others were only altered at the adenoma to carcinoma stage or only at the normal to carcinoma stage. Evaluation of the Agilent platform showed a high degree of repeatability (r = 0.98) and reasonable agreement with the NanoString platform. Our data suggest that miRNAs are highly dysregulated in colorectal tissue among individuals with colorectal cancer; the pattern of disruption varies by miRNA as tissue progresses from normal to adenoma to carcinoma.

Colorectal tumor molecular phenotype and miRNA: expression profiles and prognosis.

MiRNAs regulate gene expression by post-transcriptionally suppressing mRNA translation or by causing mRNA degradation. It has been proposed that unique miRNAs influence specific tumor molecular phenotype. In this paper, we test the hypotheses that miRNA expression differs by tumor molecular phenotype and that those differences may influence prognosis. Data come from population-based studies of colorectal cancer conducted in Utah and the Northern California Kaiser Permanente Medical Care Program. A total of 1893 carcinoma samples were run on the Agilent Human miRNA Microarray V19.0 containing 2006 miRNAs. We assessed differences in miRNA expression between TP53-mutated and non-mutated, KRAS-mutated and non-mutated, BRAF-mutated and non-mutated, CpG island methylator phenotype (CIMP) high and CIMP low, and microsatellite instability (MSI) and microsatellite stable (MSS) colon and rectal tumors. Using a Cox proportional hazard model we evaluated if those miRNAs differentially expressed by tumor phenotype influenced survival after adjusting for age, sex, and AJCC stage. There were 22 differentially expressed miRNAs for TP53-mutated colon tumors and 5 for TP53-mutated rectal tumors with a fold change of >1.49 (or <0.67). Additionally, 13 miRNAS were differentially expressed for KRAS-mutated rectal tumors, 8 differentially expressed miRNAs for colon CIMP high tumors, and 2 differentially expressed miRNAs for BRAF-mutated colon tumors. The majority of differentially expressed miRNAS were observed between MSI and MSS tumors (94 differentially expressed miRNAs for colon; 41 differentially expressed miRNAs for rectal tumors). Of these miRNAs differentially expressed between MSI and MSS tumors, the majority were downregulated. Ten of the differentially expressed miRNAs were associated with survival; after adjustment for MSI status, five miRNAS, miR-196b-5p, miR-31-5p, miR-99b-5p, miR-636, and miR-192-3p, were significantly associated with survival. In summary, it appears that the majority of miRNAs that are differentially expressed by tumor molecular phenotype are MSI tumors. However, these miRNAs appear to have minimal effect on prognosis.

Increased Susceptibility to Atrial Fibrillation Secondary to Atrial Fibrosis in Transgenic Goats Expressing Transforming Growth Factor-ß1.

INTRODUCTION: Large animal models of progressive atrial fibrosis would provide an attractive platform to study relationship between structural and electrical remodeling in atrial fibrillation (AF). Here we established a new transgenic goat model of AF with cardiac specific overexpression of TGF-ß1 and investigated the changes in the cardiac structure and function leading to AF. METHODS AND RESULTS: Transgenic goats with cardiac specific overexpression of constitutively active TGF-ß1 were generated by somatic cell nuclear transfer. We examined myocardial tissue, ECGs, echocardiographic data, and AF susceptibility in transgenic and wild-type control goats. Transgenic goats exhibited significant increase in fibrosis and myocyte diameters in the atria compared to controls, but not in the ventricles. P-wave duration was significantly greater in transgenic animals starting at 12 months of age, but no significant chamber enlargement was detected, suggesting conduction slowing in the atria. Furthermore, this transgenic goat model exhibited a significant increase in AF vulnerability. Six of 8 transgenic goats (75%) were susceptible to AF induction and exhibited sustained AF (>2 minutes), whereas none of 6 controls displayed sustained AF (P < 0.01). Length of induced AF episodes was also significantly greater in the transgenic group compared to controls (687 ± 212.02 seconds vs. 2.50 ± 0.88 seconds, P < 0.0001), but no persistent or permanent AF was observed. CONCLUSION: A novel transgenic goat model with a substrate for AF was generated. In this model, cardiac overexpression of TGF-ß1 led to an increase in fibrosis and myocyte size in the atria, and to progressive P-wave prolongation. We suggest that these factors underlie increased AF susceptibility.

Incorporation of subject-level covariates in quantile normalization of miRNA data.

BACKGROUND: Most currently-used normalization methods for miRNA array data are based on methods developed for mRNA arrays despite fundamental differences between the data characteristics. The application of conventional quantile normalization can mask important expression differences by ignoring demographic and environmental factors. We present a generalization of the conventional quantile normalization method, making use of available subject-level covariates in a colorectal cancer study. RESULTS: In simulation, our weighted quantile normalization method is shown to increase statistical power by as much as 10 % when relevant subject-level covariates are available. In application to the colorectal cancer study, this increase in power is also observed, and previously-reported dysregulated miRNAs are rediscovered. CONCLUSIONS: When any subject-level covariates are available, the weighted quantile normalization method should be used over the conventional quantile normalization method.

A shortcut for multiple testing on the directed acyclic graph of gene ontology.

BACKGROUND: Gene set testing has become an important analysis technique in high throughput microarray and next generation sequencing studies for uncovering patterns of differential expression of various biological processes. Often, the large number of gene sets that are tested simultaneously require some sort of multiplicity correction to account for the multiplicity effect. This work provides a substantial computational improvement to an existing familywise error rate controlling multiplicity approach (the Focus Level method) for gene set testing in high throughput microarray and next generation sequencing studies using Gene Ontology graphs, which we call the Short Focus Level. RESULTS: The Short Focus Level procedure, which performs a shortcut of the full Focus Level procedure, is achieved by extending the reach of graphical weighted Bonferroni testing to closed testing situations where restricted hypotheses are present, such as in the Gene Ontology graphs. The Short Focus Level multiplicity adjustment can perform the full top-down approach of the original Focus Level procedure, overcoming a significant disadvantage of the otherwise powerful Focus Level multiplicity adjustment. The computational and power differences of the Short Focus Level procedure as compared to the original Focus Level procedure are demonstrated both through simulation and using real data. CONCLUSIONS: The Short Focus Level procedure shows a significant increase in computation speed over the original Focus Level procedure (as much as ~15,000 times faster). The Short Focus Level should be used in place of the Focus Level procedure whenever the logical assumptions of the Gene Ontology graph structure are appropriate for the study objectives and when either no a priori focus level of interest can be specified or the focus level is selected at a higher level of the graph, where the Focus Level procedure is computationally intractable.

A graphical weighted power improving multiplicity correction approach for SNP selections.

Controlling for the multiplicity effect is an essential part of determining statistical significance in large-scale single-locus association genome scans on Single Nucleotide Polymorphisms (SNPs). Bonferroni adjustment is a commonly used approach due to its simplicity, but is conservative and has low power for large-scale tests. The permutation test, which is a powerful and popular tool, is computationally expensive and may mislead in the presence of family structure. We propose a computationally efficient and powerful multiple testing correction approach for Linkage Disequilibrium (LD) based Quantitative Trait Loci (QTL) mapping on the basis of graphical weighted-Bonferroni methods. The proposed multiplicity adjustment method synthesizes weighted Bonferroni-based closed testing procedures into a powerful and versatile graphical approach. By tailoring different priorities for the two hypothesis tests involved in LD based QTL mapping, we are able to increase power and maintain computational efficiency and conceptual simplicity. The proposed approach enables strong control of the familywise error rate (FWER). The performance of the proposed approach as compared to the standard Bonferroni correction is illustrated by simulation and real data. We observe a consistent and moderate increase in power under all simulated circumstances, among different sample sizes, heritabilities, and number of SNPs. We also applied the proposed method to a real outbred mouse HDL cholesterol QTL mapping project where we detected the significant QTLs that were highlighted in the literature, while still ensuring strong control of the FWER.

Transcriptional profiling by RNA-Seq of peri-attachment porcine embryos generated by a variety of assisted reproductive technologies.

Substantial mortality of in vitro manipulated porcine embryos is observed during peri-attachment development. Herein we describe our efforts to characterize the transcriptomes of embryonic disc (ED) and trophectoderm (TE) cells from porcine embryos derived from in vivo fertilization, in vitro fertilization (IVF), parthenogenetic oocyte activation (PA), and somatic cell nuclear transfer (SCNT) on days 10, 12, and 14 of gestation. The IVF, PA, and SCNT embryos were generated with in vitro matured oocytes and were cultured overnight in vitro before being transferred to recipient females. Sequencing of cDNA from the resulting embryonic samples was accomplished with the Genome Analyzer IIx platform from Illumina. Reads were aligned to a custom-built swine transcriptome. A generalized linear model was fit for ED and TE samples separately, accounting for embryo type, gestation day, and their interaction. Those genes with significant differences between embryo types were characterized in terms of gene ontologies and KEGG pathways. Transforming growth factor-ß signaling was downregulated in the EDs of IVF embryos. In TE cells from IVF embryos, ubiquitin-mediated proteolysis and ErbB signaling were aberrantly regulated. Expression of genes involved in chromatin modification, gene silencing by RNA, and apoptosis was significantly disrupted in ED cells from SCNT embryos. In summary, we have used high-throughput sequencing technologies to compare gene expression profiles of various embryo types during peri-attachment development. We expect that these data will provide important insight into the root causes of (and possible opportunities for mitigation of) suboptimal development of embryos derived from assisted reproductive technologies.

Preadaptation to cold stress in Salmonella enterica serovar Typhimurium increases survival during subsequent acid stress exposure.

Salmonella is an important cause of bacterial food-borne gastroenteritis. Salmonella encounters multiple abiotic stresses during pathogen elimination methods used in food processing, and these stresses may influence its subsequent survivability within the host or in the environment. Upon ingestion, Salmonella is exposed to gastrointestinal acidity, a first line of the host innate defense system. This study tested the hypothesis that abiotic stresses encountered during food processing alter the metabolic mechanisms in Salmonella that enable survival and persistence during subsequent exposure to the host gastrointestinal acidic environment. Out of the four different abiotic stresses tested, viz., cold, peroxide, osmotic, and acid, preadaptation of the log-phase culture to cold stress (5°C for 5 h) significantly enhanced survival during subsequent acid stress (pH 4.0 for 90 min). The gene expression profile of Salmonella preadapted to cold stress revealed induction of multiple genes associated with amino acid metabolism, oxidative stress, and DNA repair, while only a few of the genes in the above-mentioned stress response and repair pathways were induced upon exposure to acid stress alone. Preadaptation to cold stress decreased the NAD+/NADH ratio and hydroxyl (OH·) radical formation compared with those achieved with the exposure to acid stress alone, indicating alteration of aerobic respiration and the oxidative state of the bacteria. The results from this study suggest that preadaptation to cold stress rescues Salmonella from the deleterious effect of subsequent acid stress exposure by induction of genes involved in stress response and repair pathways, by modification of aerobic respiration, and by redox modulation.

The implementation of random survival forests in conflict management data: An examination of power sharing and third party mediation in post-conflict countries.

Time-to-event analysis is a common occurrence in political science. In recent years, there has been an increased usage of machine learning methods in quantitative political science research. This article advocates for the implementation of machine learning duration models to assist in a sound model selection process. We provide a brief tutorial introduction to the random survival forest (RSF) algorithm and contrast it to a popular predecessor, the Cox proportional hazards model, with emphasis on methodological utility for political science researchers. We implement both methods for simulated time-to-event data and the Power-Sharing Event Dataset (PSED) to assist researchers in evaluating the merits of machine learning duration models. We provide evidence of significantly higher survival probabilities for peace agreements with 3rd party mediated design and implementation. We also detect increased survival probabilities for peace agreements that incorporate territorial power-sharing and avoid multiple rebel party signatories. Further, the RSF, a previously under-used method for analyzing political science time-to event data, provides a novel approach for ranking of peace agreement criteria importance in predicting peace agreement duration. Our findings demonstrate a scenario exhibiting the interpretability and performance of RSF for political science time-to-event data. These findings justify the robust interpretability and competitive performance of the random survival forest algorithm in numerous circumstances, in addition to promoting a diverse, holistic model-selection process for time-to-event political science data.

A comparison of probe-level and probeset models for small-sample gene expression data.

BACKGROUND: Statistical methods to tentatively identify differentially expressed genes in microarray studies typically assume larger sample sizes than are practical or even possible in some settings. RESULTS: The performance of several probe-level and probeset models was assessed graphically and numerically using three spike-in datasets. Based on the Affymetrix GeneChip, a novel nested factorial model was developed and found to perform competitively on small-sample spike-in experiments. CONCLUSIONS: Statistical methods with test statistics related to the estimated log fold change tend to be more consistent in their performance on small-sample gene expression data. For such small-sample experiments, the nested factorial model can be a useful statistical tool. This method is implemented in freely-available R code (affyNFM), available with a tutorial document at http://www.stat.usu.edu/~jrstevens.