BACKGROUND: Massively parallel sequencing includes many liquid handling steps which introduce the possibility of sample swaps, mixing, and duplication. The unique profile of inherited variants in human genomes allows for comparison of sample identity using sequence data. A comparison of all samples vs. each other (all vs. all) provides both identification of mismatched samples and the possibility of resolving swapped samples. However, all vs. all comparison complexity grows as the square of the number of samples, so efficiency becomes essential. RESULTS: We have developed a tool for fast all vs. all genotype comparison using low level bitwise operations built into the Perl programming language. Importantly, we have also developed a complete workflow allowing users to start with either raw FASTQ sequence files, aligned BAM files, or genotype VCF files and automatically generate comparison metrics and summary plots. The tool is freely available at https://github.com/teerjk/TimeAttackGenComp/ . CONCLUSIONS: A fast and easy to use method for genotype comparison as described here is an important tool to ensure high quality and robust results in sequencing studies.

Pathway-level survival analysis offers the opportunity to examine molecular pathways and immune signatures that influence patient outcomes. However, available survival analysis algorithms are limited in pathway-level function and lack a streamlined analytical process. Here we present a comprehensive pathway-level survival analysis suite, PATH-SURVEYOR, which includes a Shiny user interface with extensive features for systematic exploration of pathways and covariates in a Cox proportional-hazard model. Moreover, our framework offers an integrative strategy for performing Hazard Ratio ranked Gene Set Enrichment Analysis and pathway clustering. As an example, we applied our tool in a combined cohort of melanoma patients treated with checkpoint inhibition (ICI) and identified several immune populations and biomarkers predictive of ICI efficacy. We also analyzed gene expression data of pediatric acute myeloid leukemia (AML) and performed an inverse association of drug targets with the patient's clinical endpoint. Our analysis derived several drug targets in high-risk KMT2A-fusion-positive patients, which were then validated in AML cell lines in the Genomics of Drug Sensitivity database. Altogether, the tool offers a comprehensive suite for pathway-level survival analysis and a user interface for exploring drug targets, molecular features, and immune populations at different resolutions.

MOTIVATION: Missingness in label-free mass spectrometry is inherent to the technology. A computational approach to recover missing values in metabolomics and proteomics datasets is important. Most existing methods are designed under a particular assumption, either missing at random or under the detection limit. If the missing pattern deviates from the assumption, it may lead to biased results. Hence, we investigate the missing patterns in free mass spectrometry data and develop an omnibus approach GMSimpute, to allow effective imputation accommodating different missing patterns. RESULTS: Three proteomics datasets and one metabolomics dataset indicate missing values could be a mixture of abundance-dependent and abundance-independent missingness. We assess the performance of GMSimpute using simulated data (with a wide range of 80 missing patterns) and metabolomics data from the Cancer Genome Atlas breast cancer and clear cell renal cell carcinoma studies. Using Pearson correlation and normalized root mean square errors between the true and imputed abundance, we compare its performance to K-nearest neighbors' type approaches, Random Forest, GSimp, a model-based method implemented in DanteR and minimum values. The results indicate GMSimpute provides higher accuracy in imputation and exhibits stable performance across different missing patterns. In addition, GMSimpute is able to identify the features in downstream differential expression analysis with high accuracy when applied to the Cancer Genome Atlas datasets. AVAILABILITY AND IMPLEMENTATION: GMSimpute is on CRAN: https://cran.r-project.org/web/packages/GMSimpute/index.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Many gene expression normalization algorithms exist for Affymetrix GeneChip microarrays. The most popular of these is RMA, primarily due to the precision and low noise produced during the process. A significant strength of this and similar approaches is the use of the entire set of arrays during both normalization and model-based estimation of signal. However, this leads to differing estimates of expression based on the starting set of arrays, and estimates can change when a single, additional chip is added to the set. Additionally, outlier chips can impact the signals of other arrays, and can themselves be skewed by the majority of the population. RESULTS: We developed an approach, termed IRON, which uses the best-performing techniques from each of several popular processing methods while retaining the ability to incrementally renormalize data without altering previously normalized expression. This combination of approaches results in a method that performs comparably to existing approaches on artificial benchmark datasets (i.e. spike-in) and demonstrates promising improvements in segregating true signals within biologically complex experiments. CONCLUSIONS: By combining approaches from existing normalization techniques, the IRON method offers several advantages. First, IRON normalization occurs pair-wise, thereby avoiding the need for all chips to be normalized together, which can be important for large data analyses. Secondly, the technique does not require similarity in signal distribution across chips for normalization, which can be important for maintaining biologically relevant differences in a heterogeneous background. Lastly, IRON introduces fewer post-processing artifacts, particularly in data whose behavior violates common assumptions. Thus, the IRON method provides a practical solution to common needs of expression analysis. A software implementation of IRON is available at [http://gene.moffitt.org/libaffy/].

Gene-expression microarray datasets often consist of a limited number of samples with a large number of gene-expression measurements, usually on the order of thousands. Therefore, dimensionality reduction is critical prior to any classification task. In this work, the iterative feature perturbation method (IFP), an embedded gene selector, is introduced and applied to four microarray cancer datasets: colon cancer, leukemia, Moffitt colon cancer, and lung cancer. We compare results obtained by IFP to those of support vector machine-recursive feature elimination (SVM-RFE) and the t-test as a feature filter using a linear support vector machine as the base classifier. Analysis of the intersection of gene sets selected by the three methods across the four datasets was done. Additional experiments included an initial pre-selection of the top 200 genes based on their p values. IFP and SVM-RFE were then applied on the reduced feature sets. These results showed up to 3.32% average performance improvement for IFP across the four datasets. A statistical analysis (using the Friedman/Holm test) for both scenarios showed the highest accuracies came from the t-test as a filter on experiments without gene pre-selection. IFP and SVM-RFE had greater classification accuracy after gene pre-selection. Analysis showed the t-test is a good gene selector for microarray data. IFP and SVM-RFE showed performance improvement on a reduced by t-test dataset. The IFP approach resulted in comparable or superior average class accuracy when compared to SVM-RFE on three of the four datasets. The same or similar accuracies can be obtained with different sets of genes.

Typically, thousands of gene expression levels are recorded for a group of patients, leading to the situation where the number of features far exceeds the number of examples. To combat this, researchers would want to combine gene expression data collected at different sites into one data set to reduce the magnitude of the difference between the number of features (genes) and examples (samples). This makes gene selection a critical component of any process to build models using gene expression data. For instance, in the domain of ordering cancer patients based on survival time, one might assume that utilizing genes related to cancer development and progression will allow the best model to be built. In this paper, we explore two different gene selection techniques and examine how well the genes selected compare between methods. We also check gene set consistency between data sets collected using the same protocols at different research institutions. It is shown that gene selection can result in very different sets given different training data.

Many genes and a small number of samples are problematic characteristics of microarray datasets. We investigated the impact on classification accuracy of gene selection approaches on filtered-to-200-gene datasets. Four datasets were used with 3 filters: Student's t-test, information gain, and reliefF. We applied Iterative Feature Perturbation (IFP) and Recursive Feature Elimination (SVM-RFE) for further gene selection. Both approaches resulted in accuracy improvement when used with the t-test-filtered datasets, but not when information gain or reliefF were used. An AUC analysis of the IFP and SVM-RFE accuracy curves indicated that both methods reached the highest AUC values after t-test filtering. A statistical analysis of accuracy across the best 50 genes using the Friedman/Holm test showed that IFP and SVM-RFE were significantly more accurate more often when applied to the t-test-filtered gene sets. Surprisingly, the simple t-test, applied as a filter, results in the best overall SVM accuracy and is at least as accurate as the other, more complicated filter methods.

Microarray datasets are often limited to a small number of samples with a large number of gene expressions. Therefore, dimensionality reduction through a feature/gene selection process is highly important for classification purposes. In this paper, a feature perturbation method we previously introduced is applied to do gene selection from microarray data. A publicly available colon cancer dataset is used in our experiments. In comparison with SVM-RFE, our method is better with feature sets of between 10 and 80, however for less than 10 features SVM-RFE results in higher accuracy. An analysis of the area under the curve of the feature perturbation method for the top 50 and 25 features is performed, aiming to determine the proper amount of noise to be applied. We show that a good set of small features/genes can be found using the feature perturbation method.

UNLABELLED: Peptide-based proteomics supports identification and quantification as well as localization of post-translational modifications (PTMs) within proteins extracted from biological samples. The 'bottom-up' approach involves the digestion of proteins into peptide fragments that can be detected and sequenced with liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). A web-based application, iPEP, was developed to compare the effectiveness of different proteolytic digests in detecting specific sequences. Furthermore, peptide populations can be examined to help optimize detection of certain groups of proteins relative to the proteome and the digested peptidome. The application reports proteolytic peptide sequences, theoretical molecular weights and functional annotations using Gene Ontology (GO) terms. The iPEP tool can assist with experimental design by maximizing the detection of proteins, consensus sites and modified residues of interest for individual proteins or as part of large-scale proteomic assays. AVAILABILITY: http://ipep.moffitt.org

Gene expression analysis techniques identify important genes that predict specified outcomes based on sample characteristics. Given the small sample sizes common to these studies and the large dimensionality of the data, feature selection methods are essential. In addition, cancer-related expression analysis often involves imbalanced datasets due to rare forms of disease. Popular methods of feature selection employ univariate techniques to identify the features most suitable for analysis. We propose a multivariate technique for selecting accurate subsets of features using an approach based on random subspaces. The random subspace method is used to explore random combinations of features and only subspaces that produce accurate classifiers are retained. The method is tested on two independent gene expression datasets and compared with a univariate approach. The multivariate feature selection method resulted in a 33% improvement in classification accuracy overall and 90% improvement in classification accuracy for the minority class.

SUMMARY: Affymetrix GeneChip microarrays are increasingly used in gene expression studies and in greater number. A software library was developed that supports Affymetrix file formats and implements two popular summary algorithms (MAS5.0 and RMA). The library is modular in design for integration into larger systems and processing pipelines. Additionally, a graphical interface (GENE) was developed to allow end-user access to the functionality within the library. AVAILABILITY: libaffy is free to use under the GNU GPL license. The source code and Windows binaries can be freely accessed from the website http://src.moffitt.usf.edu/libaffy. Additional API documentation and user manual are available.

Distributed machine learning can be realized using a divide and conquer methodology. One such divide and conquer method is learning from soft partitions of data. By examining the decomposition of classifier error into bias and variance terms, we see that learning from smaller partitions of data introduces higher variance. In this paper, we investigate the use of a particular variance reduction technique, randomized C4.5, when learning from soft partitions of data. This approach maintains the distributed nature of the learning algorithm while boosting the overall classification accuracy. Experiments on six machine learning datasets demonstrate the improved accuracy gains by reducing classifier variance. In particular, learning from soft partitions of data can produce more accurate classifiers than using an ensemble of randomized decision trees constructed from the entire dataset, which in turn results in a more accurate classifier than building a single decision tree.

We experimentally evaluate randomization-based approaches to creating an ensemble of decision-tree classifiers. Unlike methods related to boosting, all of the eight approaches considered here create each classifier in an ensemble independently of the other classifiers. Experiments were performed on 28 publicly available datasets, using C4.5 release 8 as the base classifier. While each of the other seven approaches has some strengths, we find that none of them is consistently more accurate than standard bagging when tested for statistical significance.

Type-1 diabetes is a life-long disorder that often occurs in children or young adults. Susceptibility to type-1 diabetes is believed to be dependent on both genetic and environmental causes. Genetic screening is becoming a common tool for evaluating risk, potentially before the onset of the disease. However, interpreting the results of genetic tests involves many rules taken from the literature on the correlation between risk and genetic markers. We introduce an expert system based approach in which the evaluation of risk for type-1 diabetes can be automated. This method can act as a tool for medical professionals and provides the crucial explanatory component, including citations to the relevant literature. The expert system was developed in conjunction with geneticists. As part of the corresponding clinical trial, the expert system correctly identified the at-risk genetic profiles in a set of 253 subjects evaluated by human experts.

Clustering is a useful approach in image segmentation, data mining, and other pattern recognition problems for which unlabeled data exist. Fuzzy clustering using fuzzy c-means or variants of it can provide a data partition that is both better and more meaningful than hard clustering approaches. The clustering process can be quite slow when there are many objects or patterns to be clustered. This paper discusses the algorithm brFCM, which is able to reduce the number of distinct patterns which must be clustered without adversely affecting the partition quality. The reduction is done by aggregating similar examples and then using a weighted exemplar in the clustering process. The reduction in the amount of clustering data allows a partition of the data to be produced faster. The algorithm is applied to the problem of segmenting 32 magnetic resonance images into different tissue types and the problem of segmenting 172 infrared images into trees, grass and target. Average speed-ups of as much as 59-290 times a traditional implementation of fuzzy c-means were obtained using brFCM, while producing partitions that are equivalent to those produced by fuzzy c-means.

Ensembles of classifiers offer promise in increasing overall classification accuracy. The availability of extremely large datasets has opened avenues for application of distributed and/or parallel learning to efficiently learn models of them. In this paper, distributed learning is done by training classifiers on disjoint subsets of the data. We examine a random partitioning method to create disjoint subsets and propose a more intelligent way of partitioning into disjoint subsets using clustering. It was observed that the intelligent method of partitioning generally performs better than random partitioning for our datasets. In both methods a significant gain in accuracy may be obtained by applying bagging to each of the disjoint subsets, creating multiple diverse classifiers. The significance of our finding is that a partition strategy for even small/moderate sized datasets when combined with bagging can yield better performance than applying a single learner using the entire dataset.