A new approach to describe the taxonomic structure of microbiome and its application to assess the relationship between microbial niches

BACKGROUND

Data from microbiomes from multiple niches is often collected, but methods to analyse these often ignore associations between niches. One interesting case is that of the oral microbiome. Its composition is receiving increasing attention due to reports on its associations with general health. While the oral cavity includes different niches, multi-niche microbiome data analysis is conducted using a single niche at a time and, therefore, ignores other niches that could act as confounding variables. Understanding the interaction between niches would assist interpretation of the results, and help improve our understanding of multi-niche microbiomes.

METHODS

In this study, we used a machine learning technique called latent Dirichlet allocation (LDA) on two microbiome datasets consisting of several niches. LDA was used on both individual niches and all niches simultaneously. On individual niches, LDA was used to decompose each niche into bacterial sub-communities unveiling their taxonomic structure. These sub-communities were then used to assess the relationship between microbial niches using the global test. On all niches simultaneously, LDA allowed us to extract meaningful microbial patterns. Sets of co-occurring operational taxonomic units (OTUs) comprising those patterns were then used to predict the original location of each sample.

RESULTS

Our approach showed that the per-niche sub-communities displayed a strong association between supragingival plaque and saliva, as well as between the anterior and posterior tongue. In addition, the LDA-derived microbial signatures were able to predict the original sample niche illustrating the meaningfulness of our sub-communities. For the multi-niche oral microbiome dataset we had an overall accuracy of 76%, and per-niche sensitivity of up to 83%. Finally, for a second multi-niche microbiome dataset from the entire body, microbial niches from the oral cavity displayed stronger associations to each other than with those from other parts of the body, such as niches within the vagina and the skin.

CONCLUSION

Our LDA-based approach produces sets of co-occurring taxa that can describe niche composition. LDA-derived microbial signatures can also be instrumental in summarizing microbiome data, for both descriptions as well as prediction.

Electronic Nose Sensor Drift Affects Diagnostic Reliability and Accuracy of Disease-Specific Algorithms

Sensor drift is a well-known disadvantage of electronic nose (eNose) technology and may affect the accuracy of diagnostic algorithms. Correction for this phenomenon is not routinely performed. The aim of this study was to investigate the influence of eNose sensor drift on the development of a disease-specific algorithm in a real-life cohort of inflammatory bowel disease patients (IBD). In this multi-center cohort, patients undergoing colonoscopy collected a fecal sample prior to bowel lavage. Mucosal disease activity was assessed based on endoscopy. Controls underwent colonoscopy for various reasons and had no endoscopic abnormalities. Fecal eNose profiles were measured using Cyranose 320®. Fecal samples of 63 IBD patients and 63 controls were measured on four subsequent days. Sensor data displayed associations with date of measurement, which was reproducible across all samples irrespective of disease state, disease activity state, disease localization and diet of participants. Based on logistic regression, corrections for sensor drift improved accuracy to differentiate between IBD patients and controls based on the significant differences of six sensors (p = 0.004; p < 0.001; p = 0.001; p = 0.028; p < 0.001 and p = 0.005) with an accuracy of 0.68. In this clinical study, short-term sensor drift affected fecal eNose profiles more profoundly than clinical features. These outcomes emphasize the importance of sensor drift correction to improve reliability and repeatability, both within and across eNose studies.

Distinct Patterns of Myeloid Cell Infiltration in Patients With hrHPV-Positive and hrHPV-Negative Penile Squamous Cell Carcinoma: The Importance of Assessing Myeloid Cell Densities Within the Spatial Context of the Tumor

Comprehensive analysis of tumor infiltrating myeloid cells in the tumor microenvironment of penile squamous cell carcinoma (PSCC) is lacking. In this retrospective study, for the first time, PSCC resection specimens (N = 103) were annotated into the following compartments: intratumoral tumor (IT Tumor), intratumoral stroma (IT Stroma), peritumoral tumor (PT Tumor) and peritumoral stroma (PT Stroma) compartments. We then quantified CD14+, CD68+ and CD163+ myeloid cells within these compartments using an image analysis software and assessed their association with various clinical parameters, including high-risk human papillomavirus (hrHPV) status. In the total cohort, hrHPV status, grade of differentiation, age and tumor size were associated with myeloid cell densities. hrHPV⁺ tumors had higher infiltration rates of CD14+, CD68+ and CD163+ myeloid cells in the IT tumor compartment (p < 0.001, for all) compared to hrHPV^- tumors. Furthermore, when examining the association between compartment-specific infiltration and differentiation grade, increased myeloid cell densities in the IT tumor compartment were associated with a more advanced histological grade (p < 0.001, for all). This association remained significant when the hrHPV^- cohort (N = 60) was analyzed (CD14+ p = 0.001; CD68+ p < 0.001; CD163+ p = 0.004). Subgroup analysis in the hrHPV⁺ group (N = 43) showed that high infiltration rates of CD68+ and CD163+ cells in the PT tumor compartment were associated with lymph node (LN) metastasis (p = 0.031 and p = 0.026, respectively). Regarding the association between myeloid cell densities and disease-specific survival, the risk of death was found to decrease slightly as the number of myeloid cells in the IT tumor compartment increased (CD14+ p = 0.04; CD68+ p = 0.05; CD163+ p = 0.02). However, after adjusting for hrHPV, no independent association between myeloid densities and disease-specific survival were found. Altogether, these findings demonstrate the importance of assessing myeloid cell densities within the spatial context of the tumor. Further studies are needed to unravel the specific phenotype of myeloid cells residing in the different compartments, their effect on clinical parameters and the impact of hrHPV on the recruitment of myeloid cell populations in PSCC.

Delta-Like Ligand-Notch1 Signaling Is Selectively Modulated by HPV16 E6 to Promote Squamous Cell Proliferation and Correlates with Cervical Cancer Prognosis

Human papillomavirus (HPV) drives high-grade intraepithelial neoplasia and cancer; for unknown reasons, this occurs most often in the cervical transformation zone. Either mutation or HPV E6-driven inhibition of Notch1 can drive neoplastic development in stratified squamous epithelia. However, the contribution of Notch1 and its Delta-like ligands (DLL) to site susceptibility remains poorly understood. Here, we map DLL1/DLL4 expression in cell populations present in normal cervical biopsies by immunofluorescence. In vitro keratinocyte 2D monolayer models, growth assays, and organotypic raft cultures were used to assess the functional role of DLL-Notch signaling in uninfected cells and its modulation by HPV16 in neoplasia. An RNA sequencing-based gene signature was used to suggest the cell of origin of 279 HPV-positive cervical carcinomas from The Cancer Genome Atlas and to relate this to disease prognosis. Finally, the prognostic impact of DLL4 expression was investigated in three independent cervical cancer patient cohorts. Three molecular cervical carcinoma subtypes were identified, with reserve cell tumors the most common and linked to relatively good prognosis. Reserve cells were characterized as DLL1-/DLL4+, a proliferative phenotype that is temporarily observed during squamous metaplasia and wound healing but appears to be sustained by HPV16 E6 in raft models of low-grade and, more prominently, high-grade neoplasia. High expression of DLL4 was associated with an increased likelihood of cervical cancer-associated death and recurrence. Taken together, DLL4-Notch1 signaling reflects a proliferative cellular state transiently present during physiologic processes but inherent to cervical reserve cells, making them strongly resemble neoplastic tissue even before HPV infection has occurred. SIGNIFICANCE: This study investigates cervical cancer cell-of-origin populations and describes a DLL-Notch1 phenotype that is associated with disease prognosis and that might help identify cells that are susceptible to HPV-induced carcinogenesis.

Expression of Oncolytic Adenovirus-Encoded RNAi Molecules Is Most Effective in a pri-miRNA Precursor Format

Oncolytic adenoviruses are being developed as new anti-cancer agents. Their efficacy can be improved by incorporating RNA interference (RNAi) molecules. RNAi molecules can be expressed in various precursor formats. The aim of this study was to determine the most effective format. To this end, we constructed three Δ24-type oncolytic adenoviruses, with human microRNA-1 (miR-1) expression cassettes in short hairpin RNA (shRNA), precursor microRNA (pre-miRNA), and primary miRNA (pri-miRNA) format, respectively. The viruses were compared for virus replication, mature miR-1 expression, and target gene silencing in cancer cells. Incorporation of the cassettes had only minor effects on virus replication. Mature miR-1 expression from the pri-miRNA format reached on average 100-fold higher levels than from the other two formats. This expression remained stable upon long-term virus propagation. Infection with the pri-miR-1-expressing virus silenced the validated miR-1 targets FOXP1 and MET. Drosha knockout almost completely abrogated mature miR-1 expression, confirming that processing of adenovirus-encoded pri-miR-1 was dependent on the host cell miRNA machinery. Using simple in vitro recombination cloning, a similar virus expressing miR-26b was made and shown to silence the validated miR-26b target PTGS2. We thus provide a platform for construction of oncolytic adenoviruses with high expression of RNAi molecules of choice.

Chemopreventive targeted treatment of head and neck precancer by Wee1 inhibition

HPV-negative head and neck squamous cell carcinomas (HNSCCs) develop in precancerous changes in the mucosal lining of the upper-aerodigestive tract. These precancerous cells contain cancer-associated genomic changes and cause primary tumors and local relapses. Therapeutic strategies to eradicate these precancerous cells are very limited. Using functional genomic screens, we identified the therapeutic vulnerabilities of premalignant mucosal cells, which are shared with fully malignant HNSCC cells. We screened 319 previously identified tumor-lethal siRNAs on a panel of cancer and precancerous cell lines as well as primary fibroblasts. In total we identified 147 tumor-essential genes including 34 druggable candidates. Of these 34, 13 were also essential in premalignant cells. We investigated the variable molecular basis of the vulnerabilities in tumor and premalignant cell lines and found indications of collateral lethality. Wee1-like kinase (WEE1) was amongst the most promising targets for both tumor and precancerous cells. All four precancerous cell lines were highly sensitive to Wee1 inhibition by Adavosertib (AZD1775), while primary keratinocytes tolerated this inhibitor. Wee1 inhibition caused induction of DNA damage during S-phase followed by mitotic failure in (pre)cancer cells. In conclusion, we uncovered Wee1 inhibition as a promising chemopreventive strategy for precancerous cells, with comparable responses as fully transformed HNSCC cells.

Neoadjuvant cisplatin and paclitaxel modulate tumor-infiltrating T cells in patients with cervical cancer

Resistance to chemotherapy is widely recognized as one of the major factors limiting therapeutic efficacy and influences clinical outcomes in patients with cancer. Many studies on various tumor types have focused on combining standard-of-care chemotherapy with immunotherapy. However, for cervical cancer, the role of neoadjuvant chemotherapy (NACT) on the local immune microenvironment is largely unexplored. We performed a pilot study on 13 primary cervical tumor samples, before and after NACT, to phenotype and enumerate tumor-infiltrating T-cell subpopulations using multiplex immunohistochemistry (CD3, CD8, FoxP3, Ki67, and Tbet) and automated co-expression analysis software. A significant decrease in proliferating (Ki67⁺) CD3⁺CD8⁻ T cells and FoxP3⁺(CD3⁺CD8⁻) regulatory T cells was observed in the tumor stroma after cisplatin and paclitaxel treatment, with increased rates of cytotoxic CD8⁺ T cells, including activated and CD8⁺Tbet⁺ T cells. No effect was observed on the number of tumor-infiltrating T cells in the cervical tumor microenvironment after treatment with cisplatin only. Therefore, we conclude that patients treated with cisplatin and paclitaxel had more tumor-infiltrating T-cell modulation than patients treated with cisplatin monotherapy. These findings enhance our understanding of the immune-modulating effect of chemotherapy and warrant future combination of the standard-of-care therapy with immunotherapy to improve clinical outcome in patients with cervical cancer.

Rscreenorm: normalization of CRISPR and siRNA screen data for more reproducible hit selection

Background

Reproducibility of hits from independent CRISPR or siRNA screens is poor. This is partly due to data normalization primarily addressing technical variability within independent screens, and not the technical differences between them.

Results

We present “rscreenorm”, a method that standardizes the functional data ranges between screens using assay controls, and subsequently performs a piecewise-linear normalization to make data distributions across all screens comparable. In simulation studies, rscreenorm reduces false positives. Using two multiple-cell lines siRNA screens, rscreenorm increased reproducibility between 27 and 62% for hits, and up to 5-fold for non-hits. Using publicly available CRISPR-Cas screen data, application of commonly used median centering yields merely 34% of overlapping hits, in contrast with rscreenorm yielding 84% of overlapping hits. Furthermore, rscreenorm yielded at most 8% discordant results, whilst median-centering yielded as much as 55%.

Conclusions

Rscreenorm yields more consistent results and keeps false positive rates under control, improving reproducibility of genetic screens data analysis from multiple cell lines.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2306-z) contains supplementary material, which is available to authorized users.

A test for detecting differential indirect trans effects between two groups of samples

Integrative analysis of copy number and gene expression data can help in understanding the cis and trans effect of copy number aberrations on transcription levels of genes involved in a pathway. To analyse how these copy number mediated gene-gene interactions differ between groups of samples we propose a new method, named dNET. Our method uses ridge regression to model the network topology involving one gene's expression level, its gene dosage and the expression levels of other genes in the network. The interaction parameters are estimated by fitting the model per gene for all samples together. However, instead of testing for differential network topology per gene, dNET tests for an overall difference in estimated parameters between two groups of samples and produces a single p-value. With the help of several simulation studies, we show that dNET can detect differential network nodes with high accuracy and low rate of false positives even in the presence of differential cis effects. We also apply dNET to publicly available TCGA cancer datasets and identify pathways where copy number mediated gene-gene interactions differ between samples with cancer stage lower than stage 3 and samples with cancer stage 3 or above.

A global × global test for testing associations between two large sets of variables

In high-dimensional omics studies where multiple molecular profiles are obtained for each set of patients, there is often interest in identifying complex multivariate associations, for example, copy number regulated expression levels in a certain pathway or in a genomic region. To detect such associations, we present a novel approach to test for association between two sets of variables. Our approach generalizes the global test, which tests for association between a group of covariates and a single univariate response, to allow high-dimensional multivariate response. We apply the method to several simulated datasets as well as two publicly available datasets, where we compare the performance of multivariate global test (G2) with univariate global test. The method is implemented in R and will be available as a part of the globaltest package in R.

A test for comparing two groups of samples when analyzing multiple omics profiles

BACKGROUND

A number of statistical models has been proposed for studying the association between gene expression and copy number data in integrated analysis. The next step is to compare association patterns between different groups of samples.

RESULTS

We propose a method, named dSIM, to find differences in association between copy number and gene expression, when comparing two groups of samples. Firstly, we use ridge regression to correct for the baseline associations between copy number and gene expression. Secondly, the global test is applied to the corrected data in order to find differences in association patterns between two groups of samples. We show that dSIM detects differences even in small genomic regions in a simulation study. We also apply dSIM to two publicly available breast cancer datasets and identify chromosome arms where copy number led gene expression regulation differs between positive and negative estrogen receptor samples. In spite of differing genomic coverage, some selected arms are identified in both datasets.

CONCLUSION

We developed a flexible and robust method for studying association differences between two groups of samples while integrating genomic data from different platforms. dSIM can be used with most types of microarray/sequencing data, including methylation and microRNA expression. The method is implemented in R and will be made part of the BioConductor package SIM.

General power and sample size calculations for high-dimensional genomic data

In the design of microarray or next-generation sequencing experiments it is crucial to choose the appropriate number of biological replicates. As often the number of differentially expressed genes and their effect sizes are small and too few replicates will lead to insufficient power to detect these. On the other hand, too many replicates unnecessary leads to high experimental costs. Power and sample size analysis can guide experimentalist in choosing the appropriate number of biological replicates. Several methods for power and sample size analysis have recently been proposed for microarray data. However, most of these are restricted to two group comparisons and require user-defined effect sizes. Here we propose a pilot-data based method for power and sample size analysis which can handle more general experimental designs and uses pilot-data to obtain estimates of the effect sizes. The method can also handle χ2 distributed test statistics which enables power and sample size calculations for a much wider class of models, including high-dimensional generalized linear models which are used, e.g., for RNA-seq data analysis. The performance of the method is evaluated using simulated and experimental data from several microarray and next-generation sequencing experiments. Furthermore, we compare our proposed method for estimation of the density of effect sizes from pilot data with a recent proposed method specific for two group comparisons.

Fused lasso algorithm for Cox' proportional hazards and binomial logit models with application to copy number profiles

This paper presents an efficient algorithm based on the combination of Newton Raphson and Gradient Ascent, for using the fused lasso regression method to construct a genome-based classifier. The characteristic structure of copy number data suggests that feature selection should take genomic location into account for producing more interpretable results for genome-based classifiers. The fused lasso penalty, an extension of the lasso penalty, encourages sparsity of the coefficients and their differences by penalizing the L1-norm for both of them at the same time, thus using genomic location. The major advantage of the algorithm over other existing fused lasso optimization techniques is its ability to predict binomial as well as survival response efficiently. We apply our algorithm to two publicly available datasets in order to predict survival and binary outcomes.

Analysis of small-sample clinical genomics studies using multi-parameter shrinkage: application to high-throughput RNA interference screening

High-throughput (HT) RNA interference (RNAi) screens are increasingly used for reverse genetics and drug discovery. These experiments are laborious and costly, hence sample sizes are often very small. Powerful statistical techniques to detect siRNAs that potentially enhance treatment are currently lacking, because they do not optimally use the amount of data in the other dimension, the feature dimension. We introduce ShrinkHT, a Bayesian method for shrinking multiple parameters in a statistical model, where 'shrinkage' refers to borrowing information across features. ShrinkHT is very flexible in fitting the effect size distribution for the main parameter of interest, thereby accommodating skewness that naturally occurs when siRNAs are compared with controls. In addition, it naturally down-weights the impact of nuisance parameters (e.g. assay-specific effects) when these tend to have little effects across siRNAs. We show that these properties lead to better ROC-curves than with the popular limma software. Moreover, in a 3 + 3 treatment vs control experiment with 'assay' as an additional nuisance factor, ShrinkHT is able to detect three (out of 960) significant siRNAs with stronger enhancement effects than the positive control. These were not detected by limma. In the context of gene-targeted (conjugate) treatment, these are interesting candidates for further research.

Targeting JNK-interacting-protein-1 (JIP1) sensitises osteosarcoma to doxorubicin

Osteosarcoma (OS) is the most common primary malignant bone tumour in children and adolescents. Despite aggressive therapy, survival outcomes remain unsatisfactory, especially for patients with metastatic disease or patients with a poor chemotherapy response. Chemoresistance contributes to treatment failure. To increase the efficacy of conventional chemotherapy, essential survival pathways should be targeted concomitantly. Here, we performed a loss-of-function siRNA screen of the human kinome in SaOS-2 cells to identify critical survival kinases after doxorubicin treatment. Gene silencing of JNK-interacting-protein-1 (JIP1) elicited the most potent sensitisation to doxorubicin. This candidate was further explored as potential target for chemosensitisation in OS. A panel of OS cell lines and human primary osteoblasts was examined for sensitisation to doxorubicin using small molecule JIP1-inhibitor BI-78D3. JIP1 expression and JIP1-inhibitor effects on JNK-signalling were investigated by Western blot analysis. JIP1 expression in human OS tumours was assessed by immunohistochemistry on tissue micro arrays. BI-78D3 blocked JNK-signalling and sensitised three out of four tested OS cell lines, but not healthy osteoblasts, to treatment with doxorubicin. Combination treatment increased the induction of apoptosis. JIP1 was found to be expressed in two-thirds of human primary OS tissue samples. Patients with JIP1 positive tumours showed a trend to inferior overall survival. Collectively, JIP1 appears a clinically relevant novel target in OS to enhance the efficacy of doxorubicin treatment by means of RNA interference or pharmacological inhibition.

MicroRNA profiling can classify acute leukemias of ambiguous lineage as either acute myeloid leukemia or acute lymphoid leukemia

PURPOSE

Classification of acute leukemia is based on the commitment of leukemic cells to the myeloid or the lymphoid lineage. However, a small percentage of acute leukemia cases lack straightforward immunophenotypical lineage commitment. These leukemias of ambiguous lineage represent a heterogeneous category of acute leukemia that cannot be classified as either acute myeloid leukemia (AML) or acute lymphoid leukemia (ALL). The lack of clear classification of acute leukemias of ambiguous lineage as either AML or ALL is a hurdle in treatment choice for these patients.

EXPERIMENTAL DESIGN

Here, we compared the microRNA (miRNA) expression profiles of 17 cases with acute leukemia of ambiguous lineage and 16 cases of AML, B-cell acute lymphoid leukemia (B-ALL), and T-cell acute lymphoid leukemia (T-ALL).

RESULTS

We show that leukemias of ambiguous lineage do not segregate as a separate entity but exhibit miRNA expression profiles similar to AML, B-ALL, or T-ALL. We show that by using only 5 of the most lineage-discriminative miRNAs, we are able to define acute leukemia of ambiguous lineage as either AML or ALL.

CONCLUSION

Our results indicate the presence of a myeloid or lymphoid lineage-specific genotype, as reflected by miRNA expression, in these acute leukemias despite their ambiguous immunophenotype. miRNA-based classification of acute leukemia of ambiguous lineage might be of additional value in therapeutic decision making.

Matching of array CGH and gene expression microarray features for the purpose of integrative genomic analyses

Background

An increasing number of genomic studies interrogating more than one molecular level is published. Bioinformatics follows biological practice, and recent years have seen a surge in methodology for the integrative analysis of genomic data. Often such analyses require knowledge of which elements of one platform link to those of another. Although important, many integrative analyses do not or insufficiently detail the matching of the platforms.

Results

We describe, illustrate and discuss six matching procedures. They are implemented in the R-package sigaR (available from Bioconductor). The principles underlying the presented matching procedures are generic, and can be combined to form new matching approaches or be applied to the matching of other platforms. Illustration of the matching procedures on a variety of data sets reveals how the procedures differ in the use of the available data, and may even lead to different results for individual genes.

Conclusions

Matching of data from multiple genomics platforms is an important preprocessing step for many integrative bioinformatic analysis, for which we present six generic procedures, both old and new. They have been implemented in the R-package sigaR, available from Bioconductor.

mRNA degradation controls differentiation state-dependent differences in transcript and splice variant abundance

Expression profiling experiments usually provide a static snapshot of messenger RNA (mRNA) levels. Improved understanding of the dynamics of mRNA synthesis and degradation will aid the development of sound bioinformatic models for control of gene expression. We studied mRNA stability in proliferating and differentiated myogenic cells using whole-genome exon arrays and reported the decay rates (half life) for ∼7000 mRNAs. We showed that the stability of many mRNAs strongly depends on the differentiation status and contributes to differences in abundance of these mRNAs. In addition, alternative splicing turns out to be coupled to mRNA degradation. Although different splice forms may be produced at comparable levels, their relative abundance is partly determined by their different stabilities in proliferating and differentiated cells. Where the 3'-untranslated region (3'-UTR) was previously thought to contain most RNA stabilizing and destabilizing elements, we showed that this also holds for transcript isoforms sharing the same 3'-UTR. There are two splice variants in Itga7, of which the isoform with an extra internal exon is highly stable in differentiated cells but preferentially degraded in the cytoplasm of proliferating cells. In conclusion, control of stability and degradation emerge as important determinants for differential expression of mRNA transcripts and splice variants.

Deep sequencing-based expression analysis shows major advances in robustness, resolution and inter-lab portability over five microarray platforms

The hippocampal expression profiles of wild-type mice and mice transgenic for δC-doublecortin-like kinase were compared with Solexa/Illumina deep sequencing technology and five different microarray platforms. With Illumina's digital gene expression assay, we obtained ∼2.4 million sequence tags per sample, their abundance spanning four orders of magnitude. Results were highly reproducible, even across laboratories. With a dedicated Bayesian model, we found differential expression of 3179 transcripts with an estimated false-discovery rate of 8.5%. This is a much higher figure than found for microarrays. The overlap in differentially expressed transcripts found with deep sequencing and microarrays was most significant for Affymetrix. The changes in expression observed by deep sequencing were larger than observed by microarrays or quantitative PCR. Relevant processes such as calmodulin-dependent protein kinase activity and vesicle transport along microtubules were found affected by deep sequencing but not by microarrays. While undetectable by microarrays, antisense transcription was found for 51% of all genes and alternative polyadenylation for 47%. We conclude that deep sequencing provides a major advance in robustness, comparability and richness of expression profiling data and is expected to boost collaborative, comparative and integrative genomics studies.

Cross-species comparison of human and mouse intestinal polyps reveals conserved mechanisms in adenomatous polyposis coli (APC)-driven tumorigenesis

Expression profiling is a well established tool for the genome-wide analysis of human cancers. However, the high sensitivity of this approach combined with the well known cellular and molecular heterogeneity of cancer often result in extremely complex expression signatures that are difficult to interpret functionally. The majority of sporadic colorectal cancers are triggered by mutations in the adenomatous polyposis coli (APC) tumor suppressor gene, leading to the constitutive activation of the Wnt/beta-catenin signaling pathway and formation of adenomas. Despite this common genetic basis, colorectal cancers are very heterogeneous in their degree of differentiation, growth rate, and malignancy potential. Here, we applied a cross-species comparison of expression profiles of intestinal polyps derived from hereditary colorectal cancer patients carrying APC germline mutations and from mice carrying a targeted inactivating mutation in the mouse homologue Apc. This comparative approach resulted in the establishment of a conserved signature of 166 genes that were differentially expressed between adenomas and normal intestinal mucosa in both species. Functional analyses of the conserved genes revealed a general increase in cell proliferation and the activation of the Wnt/beta-catenin signaling pathway. Moreover, the conserved signature was able to resolve expression profiles from hereditary polyposis patients carrying APC germline mutations from those with bi-allelic inactivation of the MYH gene, supporting the usefulness of such comparisons to discriminate among patients with distinct genetic defects.

Can subtle changes in gene expression be consistently detected with different microarray platforms?

Background

The comparability of gene expression data generated with different microarray platforms is still a matter of concern. Here we address the performance and the overlap in the detection of differentially expressed genes for five different microarray platforms in a challenging biological context where differences in gene expression are few and subtle.

Results

Gene expression profiles in the hippocampus of five wild-type and five transgenic δC-doublecortin-like kinase mice were evaluated with five microarray platforms: Applied Biosystems, Affymetrix, Agilent, Illumina, LGTC home-spotted arrays. Using a fixed false discovery rate of 10% we detected surprising differences between the number of differentially expressed genes per platform. Four genes were selected by ABI, 130 by Affymetrix, 3,051 by Agilent, 54 by Illumina, and 13 by LGTC. Two genes were found significantly differentially expressed by all platforms and the four genes identified by the ABI platform were found by at least three other platforms. Quantitative RT-PCR analysis confirmed 20 out of 28 of the genes detected by two or more platforms and 8 out of 15 of the genes detected by Agilent only. We observed improved correlations between platforms when ranking the genes based on the significance level than with a fixed statistical cut-off. We demonstrate significant overlap in the affected gene sets identified by the different platforms, although biological processes were represented by only partially overlapping sets of genes. Aberrances in GABA-ergic signalling in the transgenic mice were consistently found by all platforms.

Conclusion

The different microarray platforms give partially complementary views on biological processes affected. Our data indicate that when analyzing samples with only subtle differences in gene expression the use of two different platforms might be more attractive than increasing the number of replicates. Commercial two-color platforms seem to have higher power for finding differentially expressed genes between groups with small differences in expression.

Gene expression profiling highlights defective myogenesis in DMD patients and a possible role for bone morphogenetic protein 4

Duchenne Muscular Dystrophy (DMD) is characterized by progressive muscle weakness and wasting. Despite the sustained presence of satellite cells in their skeletal muscles, muscle regeneration in DMD patients seems inefficient and unable to compensate for the continuous muscle fiber loss. To find a molecular explanation, we compared the gene expression profiles of myoblasts from healthy individuals and DMD patients during activation and differentiation in culture. DMD cultures showed significant gene expression changes, even before dystrophin is expressed. We found a higher expression level of bone morphogenetic protein 4 (BMP4) in DMD cultures, which we demonstrate to inhibit differentiation into myotubes. In the later stages of differentiation, we observed a significant decline in expression of sarcomeric genes in the absence of dystrophin, probably contributing to sarcomeric instability. These results support the hypothesis that inefficient muscle regeneration is caused by impaired myoblast differentiation and impaired maintenance of the myotubes.

The expression of 70 apoptosis genes in relation to lineage, genetic subtype, cellular drug resistance, and outcome in childhood acute lymphoblastic leukemia

Childhood acute lymphoblastic leukemia (ALL) consists of various subtypes that respond differently to cytotoxic drugs and therefore have a markedly different clinical outcome. We used microarrays to investigate, in 190 children with ALL at initial diagnosis, whether 70 key apoptosis genes were differentially expressed between leukemic subgroups defined by lineage, genetic subtype, in vitro drug resistance, and clinical outcome. The expression of 44 of 70 genes was significantly different in T-versus B-lineage ALL, 22 genes differed in hyperdiploid versus nonhyperdiploid, 16 in TEL-AML1-positive versus-negative, and 13 in E2A-rearranged versus germ-line B-lineage ALL. Expression of MCL1 and DAPK1 was significantly associated with prednisolone sensitivity, whereas BCL2L13, HRK, and TNF were related to L-asparaginase resistance. BCL2L13 overexpression was also associated with unfavorable clinical outcome (P < .001). Multivariate analysis including known risk factors revealed that BCL2L13 expression was an independent prognostic factor (P = .011). The same trend was observed in a validation group of 92 children with ALL treated on a different protocol at St Jude (P = .051). In conclusion, ALL subtypes have a unique expression pattern of apoptosis genes and our data suggest that selective genes are linked to cellular drug resistance and prognosis in childhood B-lineage ALL.

Dissecting systems-wide data using mixture models: application to identify affected cellular processes

Background

Functional analysis of data from genome-scale experiments, such as microarrays, requires an extensive selection of differentially expressed genes. Under many conditions, the proportion of differentially expressed genes is considerable, making the selection criteria a balance between the inclusion of false positives and the exclusion of false negatives.

Results

We developed an analytical method to determine a p-value threshold from a microarray experiment that is dependent on the quality and design of the data set. To this aim, populations of p-values are modeled as mathematical functions in which the parameters to describe these functions are estimated in an unsupervised manner. The strength of the method is exemplified by its application to a published gene expression data set of sporadic and familial breast tumors with BRCA1 or BRCA2 mutations.

Conclusion

We present an objective and unsupervised way to set thresholds adapted to the quality and design of the experiment. The resulting mathematical description of the data sets of genome-scale experiments enables a probabilistic approach in systems biology.

Quantile smoothing of array CGH data

MOTIVATION

Plots of array Comparative Genomic Hybridization (CGH) data often show special patterns: stretches of constant level (copy number) with sharp jumps between them. There can also be much noise. Classic smoothing algorithms do not work well, because they introduce too much rounding. To remedy this, we introduce a fast and effective smoothing algorithm based on penalized quantile regression. It can compute arbitrary quantile curves, but we concentrate on the median to show the trend and the lower and upper quartile curves showing the spread of the data. Two-fold cross-validation is used for optimizing the weight of the penalties.

RESULTS

Simulated data and a published dataset are used to show the capabilities of the method to detect the segments of changed copy numbers in array CGH data.

Genomic profiling by DNA amplification of laser capture microdissected tissues and array CGH

Comparative genomic hybridization by means of BAC microarrays (array CGH) allows high-resolution profiling of copy-number aberrations in tumor DNA. However, specific genetic lesions associated with small but clinically relevant tumor areas may pass undetected due to intra-tumor heterogeneity and/or the presence of contaminating normal cells. Here, we show that the combination of laser capture microdissection, phi29 DNA polymerase-mediated isothermal genomic DNA amplification, and array CGH allows genomic profiling of very limited numbers of cells. Moreover, by means of simple statistical models, we were able to bypass the exclusion of amplification distortions and variability prone areas, and to detect tumor-specific chromosomal gains and losses. We applied this new combined experimental and analytical approach to the genomic profiling of colorectal adenomatous polyps and demonstrated our ability to accurately detect single copy gains and losses affecting either whole chromosomes or small genomic regions from as little as 2 ng of DNA or 1000 microdissected cells.

Gene expression variation between mouse inbred strains

BACKGROUND

In this study, we investigated the effect of genetic background on expression profiles. We analysed the transcriptome of mouse hindlimb muscle of five frequently used mouse inbred strains using spotted oligonucleotide microarrays.

RESULTS

Through ANOVA analysis with a false discovery rate of 10%, we show that 1.4% of the analysed genes is significantly differentially expressed between these mouse strains. Differential expression of several of these genes has been confirmed by quantitative RT-PCR. The number of genes affected by genetic background is approximately ten-fold lower than the number of differentially expressed genes caused by a dystrophic genetic defect.

CONCLUSIONS

We conclude that evaluation of the effect of background on gene expression profiles in the tissue under study is an effective and sensible approach when comparing expression patterns in animal models with heterogeneous genetic backgrounds. Genes affected by the genetic background can be excluded in subsequent analyses of the disease-related changes in expression profiles. This is often a more effective strategy than backcrossing and inbreeding to obtain isogenic backgrounds.

Intensity-based analysis of two-colour microarrays enables efficient and flexible hybridization designs

In two-colour microarrays, the ratio of signal intensities of two co-hybridized samples is used as a relative measure of gene expression. Ratio-based analysis becomes complicated and inefficient in multi-class comparisons. We therefore investigated the validity of an intensity-based analysis procedure. To this end, two different cRNA targets were hybridized together, separately, with a common reference and in a self-self fashion on spotted 65mer oligonucleotide microarrays. We found that the signal intensity of the cRNA targets was not influenced by the presence of a target labelled in the opposite colour. This indicates that targets do not compete for binding sites on the array, which is essential for intensity-based analysis. It is demonstrated that, for good-quality arrays, the correlation of signal intensity measurements between the different hybridization designs is high (R > 0.9). Furthermore, ratio calculations from ratio- and intensity-based analyses correlated well (R > 0.8). Based on these results, we advocate the use of separate intensities rather than ratios in the analysis of two-colour long-oligonucleotide microarrays. Intensity-based analysis makes microarray experiments more efficient and more flexible: It allows for direct comparisons between all hybridized samples, while circumventing the need for a reference sample that occupies half of the hybridization capacity.

Microarray data analysis: a hierarchical T-test to handle heteroscedasticity

The analysis of differential gene expression in microarray experiments requires the development of adequate statistical tools. This article describes a simple statistical method for detecting differential expression between two conditions with a low number of replicates. When comparing two group means using a traditional t-test, gene-specific variance estimates are unstable and can lead to wrong conclusions. We construct a likelihood ratio test while modelling these variances hierarchically across all genes, and express it as a t-test statistic. By borrowing information across genes we can take advantage of their large numbers, and still yield a gene-specific test statistic. We show that this hierarchical t-test is more powerful than its traditional version and generates less false positives in a simulation study, especially with small sample sizes. This approach can be extended to cases where there are more than two groups.