Blaz Zupan

Rule Evaluation Measures: A Unifying View

Numerous measures are used for performance evaluation in machine learning. In predictive knowledge discovery, the most frequently used measure is classification accuracy. With new tasks being addressed in knowledge discovery, new measures appear. In descriptive knowledge discovery, where induced rules are not primarily intended for classification, new measures used are novelty in clausal and subgroup discovery, and support and confidence in association rule learning. Additional measures are needed as many descriptive knowledge discovery tasks involve the induction of a large set of redundant rules and the problem is the ranking and filtering of the induced rule set. In this paper we develop a unifying view on some of the existing measures for predictive and descriptive induction. We provide a common terminology and notation by means of contingency tables. We demonstrate how to trade off these measures, by using what we call weighted relative accuracy. The paper furthermore demonstrates that many rule evaluation measures developed for predictive knowledge discovery can be adapted to descriptive knowledge discovery tasks.

iCLIP Predicts the Dual Splicing Effects of TIA-RNA Interactions

Transcriptome-wide analysis of protein-RNA interactions predicts the dual splicing effects of TIA proteins, showing that their local enhancing function is associated with diverse distal splicing silencing effects.

Microarray data mining with visual programming

UNLABELLED Visual programming offers an intuitive means of combining known analysis and visualization methods into powerful applications. The system presented here enables users who are not programmers to manage microarray and genomic data flow and to customize their analyses by combining common data analysis tools to fit their needs. AVAILABILITY http://www.ailab.si/supp/bi-visprog SUPPLEMENTARY INFORMATION http://www.ailab.si/supp/bi-visprog.

iCLIP - Transcriptome-wide Mapping of Protein-RNA Interactions with Individual Nucleotide Resolution

The unique composition and spatial arrangement of RNA-binding proteins (RBPs) on a transcript guide the diverse aspects of post-transcriptional regulation1. Therefore, an essential step towards understanding transcript regulation at the molecular level is to gain positional information on the binding sites of RBPs2. Protein-RNA interactions can be studied using biochemical methods, but these approaches do not address RNA binding in its native cellular context. Initial attempts to study protein-RNA complexes in their cellular environment employed affinity purification or immunoprecipitation combined with differential display or microarray analysis (RIP-CHIP)3-5. These approaches were prone to identifying indirect or non-physiological interactions6. In order to increase the specificity and positional resolution, a strategy referred to as CLIP (UV cross-linking and immunoprecipitation) was introduced7,8. CLIP combines UV cross-linking of proteins and RNA molecules with rigorous purification schemes including denaturing polyacrylamide gel electrophoresis. In combination with high-throughput sequencing technologies, CLIP has proven as a powerful tool to study protein-RNA interactions on a genome-wide scale (referred to as HITS-CLIP or CLIP-seq)9,10. Recently, PAR-CLIP was introduced that uses photoreactive ribonucleoside analogs for cross-linking11,12. Despite the high specificity of the obtained data, CLIP experiments often generate cDNA libraries of limited sequence complexity. This is partly due to the restricted amount of co-purified RNA and the two inefficient RNA ligation reactions required for library preparation. In addition, primer extension assays indicated that many cDNAs truncate prematurely at the crosslinked nucleotide13. Such truncated cDNAs are lost during the standard CLIP library preparation protocol. We recently developed iCLIP (individual-nucleotide resolution CLIP), which captures the truncated cDNAs by replacing one of the inefficient intermolecular RNA ligation steps with a more efficient intramolecular cDNA circularization (Figure 1)14. Importantly, sequencing the truncated cDNAs provides insights into the position of the cross-link site at nucleotide resolution. We successfully applied iCLIP to study hnRNP C particle organization on a genome-wide scale and assess its role in splicing regulation14.

Conserved developmental transcriptomes in evolutionarily divergent species

BackgroundEvolutionarily divergent organisms often share developmental anatomies despite vast differences between their genome sequences. The social amoebae Dictyostelium discoideum and Dictyostelium purpureum have similar developmental morphologies although their genomes are as divergent as those of man and jawed fish.ResultsHere we show that the anatomical similarities are accompanied by extensive transcriptome conservation. Using RNA sequencing we compared the abundance and developmental regulation of all the transcripts in the two species. In both species, most genes are developmentally regulated and the greatest expression changes occur during the transition from unicellularity to multicellularity. The developmental regulation of transcription is highly conserved between orthologs in the two species. In addition to timing of expression, the level of mRNA production is also conserved between orthologs and is consistent with the intuitive notion that transcript abundance correlates with the amount of protein required. Furthermore, the conservation of transcriptomes extends to cell-type specific expression.ConclusionsThese findings suggest that developmental programs are remarkably conserved at the transcriptome level, considering the great evolutionary distance between the genomes. Moreover, this transcriptional conservation may be responsible for the similar developmental anatomies of Dictyostelium discoideum and Dictyostelium purpureum.

Polymorphic members of the lag gene family mediate kin discrimination in Dictyostelium.

Self and kin discrimination are observed in most kingdoms of life and are mediated by highly polymorphic plasma membrane proteins. Sequence polymorphism, which is essential for effective recognition, is maintained by balancing selection. Dictyostelium discoideum are social amoebas that propagate as unicellular organisms but aggregate upon starvation and form fruiting bodies with viable spores and dead stalk cells. Aggregative development exposes Dictyostelium to the perils of chimerism, including cheating, which raises questions about how the victims survive in nature and how social cooperation persists. Dictyostelids can minimize the cost of chimerism by preferential cooperation with kin, but the mechanisms of kin discrimination are largely unknown. Dictyostelium lag genes encode transmembrane proteins with multiple immunoglobulin (Ig) repeats that participate in cell adhesion and signaling. Here, we describe their role in kin discrimination. We show that lagB1 and lagC1 are highly polymorphic in natural populations and that their sequence dissimilarity correlates well with wild-strain segregation. Deleting lagB1 and lagC1 results in strain segregation in chimeras with wild-type cells, whereas elimination of the nearly invariant homolog lagD1 has no such consequences. These findings reveal an early evolutionary origin of kin discrimination and provide insight into the mechanism of social recognition and immunity.

Intelligent Data Analysis in Medicine and Pharmacology

From the Publisher: Intelligent Data Analysis in Medicine and Pharmacology consists of selected (and thoroughly revised) papers presented at the First International Workshop on Intelligent Data Analysis in Medicine and Pharmacology (IDAMAP-96) held in Budapest in August 1996 as part of the 12th European Conference on Artificial Intelligence (ECAI-96). IDAMAP-96 was organized with the motivation to gather scientists and practitioners interested in computational data analysis methods applied to medicine and pharmacology, aimed at narrowing the increasing gap between excessive amounts of data stored in medical and pharmacological databases on the one hand, and the interpretation, understanding and effective use of stored data on the other hand. Besides the revised Workshop papers, the book contains a selection of contributions by invited authors. The expected readership of Intelligent Data Analysis in Medicine and Pharmacology is researchers and practitioners interested in intelligent data analysis, data mining, and knowledge discovery in databases, particularly those who are interested in using these technologies in medicine and pharmacology. Researchers and students in artificial intelligence and statistics should find this book of interest as well. Finally, much of the presented material will be interesting to physicians and pharmacologists challenged by new computational technologies, or simply in need of effectively utilizing the overwhelming volumes of data collected as a result of improved computer support in their daily professional practice.

Epistasis analysis with global transcriptional phenotypes

Classical epistasis analysis can determine the order of function of genes in pathways using morphological, biochemical and other phenotypes. It requires knowledge of the pathways phenotypic output and a variety of experimental expertise and so is unsuitable for genome-scale analysis. Here we used microarray profiles of mutants as phenotypes for epistasis analysis. Considering genes that regulate activity of protein kinase A in Dictyostelium, we identified known and unknown epistatic relationships and reconstructed a genetic network with microarray phenotypes alone. This work shows that microarray data can provide a uniform, quantitative tool for large-scale genetic network analysis.

Nomograms for visualization of naive Bayesian classifier

Besides good predictive performance, the naive Bayesian classifier can also offer a valuable insight into the structure of the training data and effects of the attributes on the class probabilities. This structure may be effectively revealed through visualization of the classifier. We propose a new way to visualize the naive Bayesian model in the form of a nomogram. The advantages of the proposed method are simplicity of presentation, clear display of the effects of individual attribute values, and visualization of confidence intervals. Nomograms are intuitive and when used for decision support can provide a visual explanation of predicted probabilities. And finally, with a nomogram, a naive Bayesian model can be printed out and used for probability prediction without the use of computer or calculator.

dictyBase—a Dictyostelium bioinformatics resource update

dictyBase (http://dictybase.org) is the model organism database for Dictyostelium discoideum. It houses the complete genome sequence, ESTs and the entire body of literature relevant to Dictyostelium. This information is curated to provide accurate gene models and functional annotations, with the goal of fully annotating the genome. This dictyBase update describes the annotations and features implemented since 2006, including improved strain and phenotype representation, integration of predicted transcriptional regulatory elements, protein domain information, biochemical pathways, improved searching and a wiki tool that allows members of the research community to provide annotations.