IQRray, a new method for Affymetrix microarray quality control, and the homologous organ conservation score, a new benchmark method for quality control metrics
Marta Rosikiewicz, Marc Robinson-Rechavi
(Submitted on 8 Oct 2013)

Motivation: Microarray results accumulated in public repositories are widely re-used in meta-analytical studies and secondary databases. The quality of the data obtained with this technology varies from experiment to experiment and efficient method for quality assessment is neces-sary to ensure their reliability. Results: The lack of a good benchmark has hampered evaluation of existing methods for quality control. In this study we propose a new inde-pendent quality metric that is based on evolutionary conservation of expression profiles. We show, using 11 large organ-specific datasets, that IQRray, a new quality metrics developed by us, exhibits the highest correlation with this reference metric, among 14 metrics tested. IQRray outperforms other methods in identification of poor quality arrays in dataset composed of arrays from many independent experiments. In con-trast, the performance of methods designed for detecting outliers in a single experiment like NUSE and RLE was low because of the inability of these method to detect datasets containing only low quality arrays, and the fact that the scores cannot be directly compared between ex-periments. Availability: The R implementation of IQRray is available at: this ftp URL

A Comparative Analysis of Ensemble Classifiers: Case Studies in Genomics
Sean Whalen, Gaurav Pandey
(Submitted on 19 Sep 2013)

The combination of multiple classifiers using ensemble methods is increasingly important for making progress in a variety of difficult prediction problems. We present a comparative analysis of several ensemble methods through two case studies in genomics, namely the prediction of genetic interactions and protein functions, to demonstrate their efficacy on real-world datasets and draw useful conclusions about their behavior. These methods include simple aggregation, meta-learning, cluster-based meta-learning, and ensemble selection using heterogeneous classifiers trained on resampled data to improve the diversity of their predictions. We present a detailed analysis of these methods across 4 genomics datasets and find the best of these methods offer statistically significant improvements over the state of the art in their respective domains. In addition, we establish a novel connection between ensemble selection and meta-learning, demonstrating how both of these disparate methods establish a balance between ensemble diversity and performance.

These are not the k-mers you are looking for: efficient online k-mer counting using a probabilistic data structure
Qingpeng Zhang, Jason Pell, Rosangela Canino-Koning, Adina Chuang Howe, C. Titus Brown
(Submitted on 11 Sep 2013)

K-mer abundance analysis is widely used for many purposes in sequence analysis, including data preprocessing for de novo assembly, repeat detection, and sequencing coverage estimation. We present the khmer software package for fast and memory efficient online counting of k-mers in sequencing data sets. Unlike previous methods based on data structures such as hash tables, suffix arrays, and trie structures, khmer relies entirely on a simple probabilistic data structure, a CountMin Sketch. The CountMin Sketch permits online updating and retrieval of k-mer counts in memory which is necessary to support streaming k-mer analysis algorithms. On sparse data sets this data structure is considerably more memory efficient than any exact data structure. In exchange, the use of a CountMin Sketch introduces a systematic overcount for k-mers; moreover, only the counts, and not the k-mers, are stored. Here we analyze the speed, the memory usage, and the miscount rate of khmer for generating k-mer frequency distributions and retrieving k-mer counts for individual k-mers. We also compare the performance of khmer to several other k-mer counting packages, including Tallymer, Jellyfish, and DSK. Finally, we examine the effectiveness of profiling sequencing error, k-mer abundance trimming, and digital normalization of reads in the context of high khmer error rates. Khmer is implemented in C++ wrapped with a Python interface, offers a tested and robust API, and is freely available under the BSD license at github.com/ged-lab/khmer.

MOSAIK: A hash-based algorithm for accurate next-generation sequencing read mapping
Wan-Ping Lee (1), Michael Stromberg (1 and 2), Alistair Ward (1), Chip Stewart (1 and 3), Erik Garrison (1), Gabor T. Marth (1) ((1) Department of Biology, Boston College, Chestnut Hill, MA, (2) Illumina, Inc., San Diego, CA, (3) Broad Institute of Harvard and Massachusetts Institute of Technology, Cambridge, MA)
(Submitted on 4 Sep 2013)

This paper presents an accurate short-read mapper for next-generation sequencing data which is widely used in the 1000 Genomes Project, and human clinical and other species genome studies.

Biological Averaging in RNA-Seq
Surojit Biswas, Yash N. Agrawal, Tatiana S. Mucyn, Jeffery L. Dangl, Corbin D. Jones
(Submitted on 3 Sep 2013)

RNA-seq has become a de facto standard for measuring gene expression. Traditionally, RNA-seq experiments are mathematically averaged — they sequence the mRNA of individuals from different treatment groups, hoping to correlate phenotype with differences in arithmetic read count averages at shared loci of interest. Alternatively, the tissue from the same individuals may be pooled prior to sequencing in what we refer to as a biologically averaged design. As mathematical averaging sequences all individuals it controls for both biological and technical variation; however, is the statistical resolution gained always worth the additional cost? To compare biological and mathematical averaging, we examined theoretical and empirical estimates of statistical efficiency and relative cost efficiency. Though less efficient at a fixed sample size, we found that biological averaging can be more cost efficient than mathematical averaging. With this motivation, we developed a differential expression classifier, ICRBC, to can detect alternatively expressed genes between biologically averaged samples. In simulation studies, we found that biological averaging and subsequent analysis with our classifier performed comparably to existing methods, such as ASC, edgeR, and DESeq, especially when individuals were pooled evenly and less than 20% of the regulome was expected to be differentially regulated. In two technically distinct mouse datasets and one plant dataset, we found that our method was over 87% concordant with edgeR for the 100 most significant features. We therefore conclude biological averaging may sufficiently control biological variation to a level that differences in gene expression may be detectable. In such situations, ICRBC can enable reliable exploratory analysis at a fraction of the cost, especially when interest lies in the most differentially expressed loci.