Precision oncology has primarily relied on coding mutations as biomarkers of response to therapies. While transcriptome analysis can provide valuable information, incorporation into workflows has been difficult. For example, the relative rather than absolute gene expression level needs to be considered, requiring differential expression analysis across samples. However, expression programs related to the cell-of-origin and tumor microenvironment effects confound the search for cancer-specific expression changes. To address these challenges, we developed an unsupervised clustering approach for discovering differential pathway expression within cancer cohorts using gene expression measurements. The hydra approach uses a Dirichlet process mixture model to automatically detect multimodally distributed genes and expression signatures without the need for matched normal tissue. We demonstrate that the hydra approach is more sensitive than widely-used gene set enrichment approaches for detecting multimodal expression signatures. Application of the hydra analysis framework to small blue round cell tumors (including rhabdomyosarcoma, synovial sarcoma, neuroblastoma, Ewing sarcoma, and osteosarcoma) identified expression signatures associated with changes in the tumor microenvironment. The hydra approach also identified an association between ATRX deletions and elevated immune marker expression in high-risk neuroblastoma. Notably, hydra analysis of all small blue round cell tumors revealed similar subtypes, characterized by changes to infiltrating immune and stromal expression signatures.

UNLABELLED: Two new tools on the UCSC Genome Browser web site provide improved ways of combining information from multiple datasets, optionally including the user's own custom track data and/or data from track hubs. The Data Integrator combines columns from multiple data tracks, showing all items from the first track along with overlapping items from the other tracks. The Variant Annotation Integrator is tailored to adding functional annotations to variant calls; it offers a more restricted set of underlying data tracks but adds predictions of each variant's consequences for any overlapping or nearby gene transcript. When available, it optionally adds additional annotations including effect prediction scores from dbNSFP for missense mutations, ENCODE regulatory summary tracks and conservation scores. AVAILABILITY AND IMPLEMENTATION: The web tools are freely available at http://genome.ucsc.edu/ and the underlying database is available for download at http://hgdownload.cse.ucsc.edu/ The software (written in C and Javascript) is available from https://genome-store.ucsc.edu/ and is freely available for academic and non-profit usage; commercial users must obtain a license. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: The study of genomic variation has provided key insights into the functional role of mutations. Predominantly, studies have focused on single nucleotide variants (SNV), which are relatively easy to detect and can be described with rich mathematical models. However, it has been observed that genomes are highly plastic, and that whole regions can be moved, removed or duplicated in bulk. These structural variants (SV) have been shown to have significant impact on phenotype, but their study has been held back by the combinatorial complexity of the underlying models. RESULTS: We describe here a general model of structural variation that encompasses both balanced rearrangements and arbitrary copy-number variants (CNV). CONCLUSIONS: In this model, we show that the space of possible evolutionary histories that explain the structural differences between any two genomes can be sampled ergodically.

UNLABELLED: Genome Browser in a Box (GBiB) is a small virtual machine version of the popular University of California Santa Cruz (UCSC) Genome Browser that can be run on a researcher's own computer. Once GBiB is installed, a standard web browser is used to access the virtual server and add personal data files from the local hard disk. Annotation data are loaded on demand through the Internet from UCSC or can be downloaded to the local computer for faster access. AVAILABILITY AND IMPLEMENTATION: Software downloads and installation instructions are freely available for non-commercial use at https://genome-store.ucsc.edu/. GBiB requires the installation of open-source software VirtualBox, available for all major operating systems, and the UCSC Genome Browser, which is open source and free for non-commercial use. Commercial use of GBiB and the Genome Browser requires a license (http://genome.ucsc.edu/license/).

MOTIVATION: Sequence mapping is the cornerstone of modern genomics. However, most existing sequence mapping algorithms are insufficiently general. RESULTS: We introduce context schemes: a method that allows the unambiguous recognition of a reference base in a query sequence by testing the query for substrings from an algorithmically defined set. Context schemes only map when there is a unique best mapping, and define this criterion uniformly for all reference bases. Mappings under context schemes can also be made stable, so that extension of the query string (e.g. by increasing read length) will not alter the mapping of previously mapped positions. Context schemes are general in several senses. They natively support the detection of arbitrary complex, novel rearrangements relative to the reference. They can scale over orders of magnitude in query sequence length. Finally, they are trivially extensible to more complex reference structures, such as graphs, that incorporate additional variation. We demonstrate empirically the existence of high-performance context schemes, and present efficient context scheme mapping algorithms. AVAILABILITY AND IMPLEMENTATION: The software test framework created for this study is available from https://registry.hub.docker.com/u/adamnovak/sequence-graphs/. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The world's genomics data will never be stored in a single repository - rather, it will be distributed among many sites in many countries. No one site will have enough data to explain genotype to phenotype relationships in rare diseases; therefore, sites must share data. To accomplish this, the genetics community must forge common standards and protocols to make sharing and computing data among many sites a seamless activity. Through the Global Alliance for Genomics and Health, we are pioneering the development of shared application programming interfaces (APIs) to connect the world's genome repositories. In parallel, we are developing an open source software stack (ADAM) that uses these APIs. This combination will create a cohesive genome informatics ecosystem. Using containers, we are facilitating the deployment of this software in a diverse array of environments. Through benchmarking efforts and big data driver projects, we are ensuring ADAM's performance and utility.

MOTIVATION: Researchers now have access to large volumes of genome sequences for comparative analysis, some generated by the plethora of public sequencing projects and, increasingly, from individual efforts. It is not possible, or necessarily desirable, that the public genome browsers attempt to curate all these data. Instead, a wealth of powerful tools is emerging to empower users to create their own visualizations and browsers. RESULTS: We introduce a pipeline to easily generate collections of Web-accessible UCSC Genome Browsers interrelated by an alignment. It is intended to democratize our comparative genomic browser resources, serving the broad and growing community of evolutionary genomicists and facilitating easy public sharing via the Internet. Using the alignment, all annotations and the alignment itself can be efficiently viewed with reference to any genome in the collection, symmetrically. A new, intelligently scaled alignment display makes it simple to view all changes between the genomes at all levels of resolution, from substitutions to complex structural rearrangements, including duplications. To demonstrate this work, we create a comparative assembly hub containing 57 Escherichia coli and 9 Shigella genomes and show examples that highlight their unique biology. AVAILABILITY AND IMPLEMENTATION: The source code is available as open source at: https://github.com/glennhickey/progressiveCactus The E.coli and Shigella genome hub is now a public hub listed on the UCSC browser public hubs Web page.

BACKGROUND: Parsimony and maximum likelihood methods of phylogenetic tree estimation and parsimony methods for genome rearrangements are central to the study of genome evolution yet to date they have largely been pursued in isolation. RESULTS: We present a data structure called a history graph that offers a practical basis for the analysis of genome evolution. It conceptually simplifies the study of parsimonious evolutionary histories by representing both substitutions and double cut and join (DCJ) rearrangements in the presence of duplications. The problem of constructing parsimonious history graphs thus subsumes related maximum parsimony problems in the fields of phylogenetic reconstruction and genome rearrangement. We show that tractable functions can be used to define upper and lower bounds on the minimum number of substitutions and DCJ rearrangements needed to explain any history graph. These bounds become tight for a special type of unambiguous history graph called an ancestral variation graph (AVG), which constrains in its combinatorial structure the number of operations required. We finally demonstrate that for a given history graph G, a finite set of AVGs describe all parsimonious interpretations of G, and this set can be explored with a few sampling moves. CONCLUSION: This theoretical study describes a model in which the inference of genome rearrangements and phylogeny can be unified under parsimony.

The UCSC Genome Browser (http://genome.ucsc.edu) is a graphical viewer for genomic data now in its 13th year. Since the early days of the Human Genome Project, it has presented an integrated view of genomic data of many kinds. Now home to assemblies for 58 organisms, the Browser presents visualization of annotations mapped to genomic coordinates. The ability to juxtapose annotations of many types facilitates inquiry-driven data mining. Gene predictions, mRNA alignments, epigenomic data from the ENCODE project, conservation scores from vertebrate whole-genome alignments and variation data may be viewed at any scale from a single base to an entire chromosome. The Browser also includes many other widely used tools, including BLAT, which is useful for alignments from high-throughput sequencing experiments. Private data uploaded as Custom Tracks and Data Hubs in many formats may be displayed alongside the rich compendium of precomputed data in the UCSC database. The Table Browser is a full-featured graphical interface, which allows querying, filtering and intersection of data tables. The Saved Session feature allows users to store and share customized views, enhancing the utility of the system for organizing multiple trains of thought. Binary Alignment/Map (BAM), Variant Call Format and the Personal Genome Single Nucleotide Polymorphisms (SNPs) data formats are useful for visualizing a large sequencing experiment (whole-genome or whole-exome), where the differences between the data set and the reference assembly may be displayed graphically. Support for high-throughput sequencing extends to compact, indexed data formats, such as BAM, bigBed and bigWig, allowing rapid visualization of large datasets from RNA-seq and ChIP-seq experiments via local hosting.

MOTIVATION: Large multiple genome alignments and inferred ancestral genomes are ideal resources for comparative studies of molecular evolution, and advances in sequencing and computing technology are making them increasingly obtainable. These structures can provide a rich understanding of the genetic relationships between all subsets of species they contain. Current formats for storing genomic alignments, such as XMFA and MAF, are all indexed or ordered using a single reference genome, however, which limits the information that can be queried with respect to other species and clades. This loss of information grows with the number of species under comparison, as well as their phylogenetic distance. RESULTS: We present HAL, a compressed, graph-based hierarchical alignment format for storing multiple genome alignments and ancestral reconstructions. HAL graphs are indexed on all genomes they contain. Furthermore, they are organized phylogenetically, which allows for modular and parallel access to arbitrary subclades without fragmentation because of rearrangements that have occurred in other lineages. HAL graphs can be created or read with a comprehensive C++ API. A set of tools is also provided to perform basic operations, such as importing and exporting data, identifying mutations and coordinate mapping (liftover). AVAILABILITY: All documentation and source code for the HAL API and tools are freely available at http://github.com/glennhickey/hal. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Identifying the cellular wiring that connects genomic perturbations to transcriptional changes in cancer is essential to gain a mechanistic understanding of disease initiation, progression and ultimately to predict drug response. We have developed a method called Tied Diffusion Through Interacting Events (TieDIE) that uses a network diffusion approach to connect genomic perturbations to gene expression changes characteristic of cancer subtypes. The method computes a subnetwork of protein-protein interactions, predicted transcription factor-to-target connections and curated interactions from literature that connects genomic and transcriptomic perturbations. RESULTS: Application of TieDIE to The Cancer Genome Atlas and a breast cancer cell line dataset identified key signaling pathways, with examples impinging on MYC activity. Interlinking genes are predicted to correspond to essential components of cancer signaling and may provide a mechanistic explanation of tumor character and suggest subtype-specific drug targets. AVAILABILITY: Software is available from the Stuart lab's wiki: https://sysbiowiki.soe.ucsc.edu/tiedie. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: A current challenge in understanding cancer processes is to pinpoint which mutations influence the onset and progression of disease. Toward this goal, we describe a method called PARADIGM-SHIFT that can predict whether a mutational event is neutral, gain-or loss-of-function in a tumor sample. The method uses a belief-propagation algorithm to infer gene activity from gene expression and copy number data in the context of a set of pathway interactions. RESULTS: The method was found to be both sensitive and specific on a set of positive and negative controls for multiple cancers for which pathway information was available. Application to the Cancer Genome Atlas glioblastoma, ovarian and lung squamous cancer datasets revealed several novel mutations with predicted high impact including several genes mutated at low frequency suggesting the approach will be complementary to current approaches that rely on the prevalence of events to reach statistical significance. AVAILABILITY: All source code is available at the github repository http:github.org/paradigmshift. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Throughout life, the cells in every individual accumulate many changes in the DNA inherited from his or her parents. Certain combinations of changes lead to cancer. During the last decade, the cost of DNA sequencing has been dropping by a factor of 10 every two years, making it now possible to read most of the three billion base genome from a patient's cancer tumor, and to try to determine all of the thousands of DNA changes in it. Under the auspices of NCI's Cancer Genome Atlas Project, 10,000 tumors will be sequenced in this manner in the next few years. Soon cancer genome sequencing will be a widespread clinical practice, and millions of tumors will be sequenced. A massive computational problem looms in interpreting these data.

BACKGROUND: Continuing research into the global multiple sequence alignment problem has resulted in more sophisticated and principled alignment methods. Unfortunately these new algorithms often require large amounts of time and memory to run, making it nearly impossible to run these algorithms on large datasets. As a solution, we present two general methods, Crumble and Prune, for breaking a phylogenetic alignment problem into smaller, more tractable sub-problems. We call Crumble and Prune meta-alignment methods because they use existing alignment algorithms and can be used with many current alignment programs. Crumble breaks long alignment problems into shorter sub-problems. Prune divides the phylogenetic tree into a collection of smaller trees to reduce the number of sequences in each alignment problem. These methods are orthogonal: they can be applied together to provide better scaling in terms of sequence length and in sequence depth. Both methods partition the problem such that many of the sub-problems can be solved independently. The results are then combined to form a solution to the full alignment problem. RESULTS: Crumble and Prune each provide a significant performance improvement with little loss of accuracy. In some cases, a gain in accuracy was observed. Crumble and Prune were tested on real and simulated data. Furthermore, we have implemented a system called Job-tree that allows hierarchical sub-problems to be solved in parallel on a compute cluster, significantly shortening the run-time. CONCLUSIONS: These methods enabled us to solve gigabase alignment problems. These methods could enable a new generation of biologically realistic alignment algorithms to be applied to real world, large scale alignment problems.

MOTIVATION: High-throughput data is providing a comprehensive view of the molecular changes in cancer tissues. New technologies allow for the simultaneous genome-wide assay of the state of genome copy number variation, gene expression, DNA methylation and epigenetics of tumor samples and cancer cell lines. Analyses of current data sets find that genetic alterations between patients can differ but often involve common pathways. It is therefore critical to identify relevant pathways involved in cancer progression and detect how they are altered in different patients. RESULTS: We present a novel method for inferring patient-specific genetic activities incorporating curated pathway interactions among genes. A gene is modeled by a factor graph as a set of interconnected variables encoding the expression and known activity of a gene and its products, allowing the incorporation of many types of omic data as evidence. The method predicts the degree to which a pathway's activities (e.g. internal gene states, interactions or high-level 'outputs') are altered in the patient using probabilistic inference. Compared with a competing pathway activity inference approach called SPIA, our method identifies altered activities in cancer-related pathways with fewer false-positives in both a glioblastoma multiform (GBM) and a breast cancer dataset. PARADIGM identified consistent pathway-level activities for subsets of the GBM patients that are overlooked when genes are considered in isolation. Further, grouping GBM patients based on their significant pathway perturbations divides them into clinically-relevant subgroups having significantly different survival outcomes. These findings suggest that therapeutics might be chosen that target genes at critical points in the commonly perturbed pathway(s) of a group of patients. AVAILABILITY: Source code available at http://sbenz.github.com/Paradigm,. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

With our ability to sequence entire genomes, we have for the first time the opportunity to compare the genomes of present day species, and deduce the trajectories by which they diversified from a common ancestral genome. For example, starting with a small shrew-like ancestor in the Cretaceous period about 100 million years ago, the different species of placental mammals radiated outward, creating a stunning diversity of forms from whales to armadillos to humans. From the genomes of present-day species, it is possible to computationally reconstruct what most of the DNA bases in the genome of the common ancestor of placental mammals must have looked like, and deduce most of the changes that lead to humans. In so doing, we discover how Darwinian evolution has shaped us at the molecular level.

MOTIVATION: Computational annotation of protein coding genes in genomic DNA is a widely used and essential tool for analyzing newly sequenced genomes. However, current methods suffer from inaccuracy and do poorly with certain types of genes. Including additional sources of evidence of the existence and structure of genes can improve the quality of gene predictions. For many eukaryotic genomes, expressed sequence tags (ESTs) are available as evidence for genes. Related genomes that have been sequenced, annotated, and aligned to the target genome provide evidence of existence and structure of genes. RESULTS: We incorporate several different evidence sources into the gene finder AUGUSTUS. The sources of evidence are gene and transcript annotations from related species syntenically mapped to the target genome using TransMap, evolutionary conservation of DNA, mRNA and ESTs of the target species, and retroposed genes. The predictions include alternative splice variants where evidence supports it. Using only ESTs we were able to correctly predict at least one splice form exactly correct in 57% of human genes. Also using evidence from other species and human mRNAs, this number rises to 77%. Syntenic mapping is well-suited to annotate genomes closely related to genomes that are already annotated or for which extensive transcript evidence is available. Native cDNA evidence is most helpful when the alignments are used as compound information rather than independent positionwise information. AVAILABILITY: AUGUSTUS is open source and available at http://augustus.gobics.de. The gene predictions for human can be browsed and downloaded at the UCSC Genome Browser (http://genome.ucsc.edu).

Correlated changes of nucleic or amino acids have provided strong information about the structures and interactions of molecules. Despite the rich literature in coevolutionary sequence analysis, previous methods often have to trade off between generality, simplicity, phylogenetic information, and specific knowledge about interactions. Furthermore, despite the evidence of coevolution in selected protein families, a comprehensive screening of coevolution among all protein domains is still lacking. We propose an augmented continuous-time Markov process model for sequence coevolution. The model can handle different types of interactions, incorporate phylogenetic information and sequence substitution, has only one extra free parameter, and requires no knowledge about interaction rules. We employ this model to large-scale screenings on the entire protein domain database (Pfam). Strikingly, with 0.1 trillion tests executed, the majority of the inferred coevolving protein domains are functionally related, and the coevolving amino acid residues are spatially coupled. Moreover, many of the coevolving positions are located at functionally important sites of proteins/protein complexes, such as the subunit linkers of superoxide dismutase, the tRNA binding sites of ribosomes, the DNA binding region of RNA polymerase, and the active and ligand binding sites of various enzymes. The results suggest sequence coevolution manifests structural and functional constraints of proteins. The intricate relations between sequence coevolution and various selective constraints are worth pursuing at a deeper level.

Taking advantage of the complete genome sequences of several mammals, we developed a novel method to detect losses of well-established genes in the human genome through syntenic mapping of gene structures between the human, mouse, and dog genomes. Unlike most previous genomic methods for pseudogene identification, this analysis is able to differentiate losses of well-established genes from pseudogenes formed shortly after segmental duplication or generated via retrotransposition. Therefore, it enables us to find genes that were inactivated long after their birth, which were likely to have evolved nonredundant biological functions before being inactivated. The method was used to look for gene losses along the human lineage during the approximately 75 million years (My) since the common ancestor of primates and rodents (the euarchontoglire crown group). We identified 26 losses of well-established genes in the human genome that were all lost at least 50 My after their birth. Many of them were previously characterized pseudogenes in the human genome, such as GULO and UOX. Our methodology is highly effective at identifying losses of single-copy genes of ancient origin, allowing us to find a few well-known pseudogenes in the human genome missed by previous high-throughput genome-wide studies. In addition to confirming previously known gene losses, we identified 16 previously uncharacterized human pseudogenes that are definitive losses of long-established genes. Among them is ACYL3, an ancient enzyme present in archaea, bacteria, and eukaryotes, but lost approximately 6 to 8 Mya in the ancestor of humans and chimps. Although losses of well-established genes do not equate to adaptive gene losses, they are a useful proxy to use when searching for such genetic changes. This is especially true for adaptive losses that occurred more than 250,000 years ago, since any genetic evidence of the selective sweep indicative of such an event has been erased.

The University of California Santa Cruz (UCSC) Known Genes dataset is constructed by a fully automated process, based on protein data from Swiss-Prot/TrEMBL (UniProt) and the associated mRNA data from Genbank. The detailed steps of this process are described. Extensive cross-references from this dataset to other genomic and proteomic data were constructed. For each known gene, a details page is provided containing rich information about the gene, together with extensive links to other relevant genomic, proteomic and pathway data. As of July 2005, the UCSC Known Genes are available for human, mouse and rat genomes. The Known Genes serves as a foundation to support several key programs: the Genome Browser, Proteome Browser, Gene Sorter and Table Browser offered at the UCSC website. All the associated data files and program source code are also available. They can be accessed at http://genome.ucsc.edu. The genomic coverage of UCSC Known Genes, RefSeq, Ensembl Genes, H-Invitational and CCDS is analyzed. Although UCSC Known Genes offers the highest genomic and CDS coverage among major human and mouse gene sets, more detailed analysis suggests all of them could be further improved.

The discoveries of microRNAs and riboswitches, among others, have shown functional RNAs to be biologically more important and genomically more prevalent than previously anticipated. We have developed a general comparative genomics method based on phylogenetic stochastic context-free grammars for identifying functional RNAs encoded in the human genome and used it to survey an eight-way genome-wide alignment of the human, chimpanzee, mouse, rat, dog, chicken, zebra-fish, and puffer-fish genomes for deeply conserved functional RNAs. At a loose threshold for acceptance, this search resulted in a set of 48,479 candidate RNA structures. This screen finds a large number of known functional RNAs, including 195 miRNAs, 62 histone 3'UTR stem loops, and various types of known genetic recoding elements. Among the highest-scoring new predictions are 169 new miRNA candidates, as well as new candidate selenocysteine insertion sites, RNA editing hairpins, RNAs involved in transcript auto regulation, and many folds that form singletons or small functional RNA families of completely unknown function. While the rate of false positives in the overall set is difficult to estimate and is likely to be substantial, the results nevertheless provide evidence for many new human functional RNAs and present specific predictions to facilitate their further characterization.

Alternative splicing contributes to both gene regulation and protein diversity. To discover broad relationships between regulation of alternative splicing and sequence conservation, we applied a systems approach, using oligonucleotide microarrays designed to capture splicing information across the mouse genome. In a set of 22 adult tissues, we observe differential expression of RNA containing at least two alternative splice junctions for about 40% of the 6,216 alternative events we could detect. Statistical comparisons identify 171 cassette exons whose inclusion or skipping is different in brain relative to other tissues and another 28 exons whose splicing is different in muscle. A subset of these exons is associated with unusual blocks of intron sequence whose conservation in vertebrates rivals that of protein-coding exons. By focusing on sets of exons with similar regulatory patterns, we have identified new sequence motifs implicated in brain and muscle splicing regulation. Of note is a motif that is strikingly similar to the branchpoint consensus but is located downstream of the 5' splice site of exons included in muscle. Analysis of three paralogous membrane-associated guanylate kinase genes reveals that each contains a paralogous tissue-regulated exon with a similar tissue inclusion pattern. While the intron sequences flanking these exons remain highly conserved among mammalian orthologs, the paralogous flanking intron sequences have diverged considerably, suggesting unusually complex evolution of the regulation of alternative splicing in multigene families.

MOTIVATION: The NCBI dbSNP database lists over 9 million single nucleotide polymorphisms (SNPs) in the human genome, but currently contains limited annotation information. SNPs that result in amino acid residue changes (nsSNPs) are of critical importance in variation between individuals, including disease and drug sensitivity. RESULTS: We have developed LS-SNP, a genomic scale software pipeline to annotate nsSNPs. LS-SNP comprehensively maps nsSNPs onto protein sequences, functional pathways and comparative protein structure models, and predicts positions where nsSNPs destabilize proteins, interfere with the formation of domain-domain interfaces, have an effect on protein-ligand binding or severely impact human health. It currently annotates 28,043 validated SNPs that produce amino acid residue substitutions in human proteins from the SwissProt/TrEMBL database. Annotations can be viewed via a web interface either in the context of a genomic region or by selecting sets of SNPs, genes, proteins or pathways. These results are useful for identifying candidate functional SNPs within a gene, haplotype or pathway and in probing molecular mechanisms responsible for functional impacts of nsSNPs. AVAILABILITY: http://www.salilab.org/LS-SNP CONTACT: [email protected] SUPPLEMENTARY INFORMATION: http://salilab.org/LS-SNP/supp-info.pdf.

Phylogenetic hidden Markov models (phylo-HMMs) have recently been proposed as a means for addressing a multi-species version of the ab initio gene prediction problem. These models allow sequence divergence, a phylogeny, patterns of substitution, and base composition all to be considered simultaneously, in a single unified probabilistic model. Here, we apply phylo-HMMs to a restricted version of the gene prediction problem in which individual exons are sought that are evolutionarily conserved across a diverse set of species. We discuss two new methods for improving prediction performance: (1) the use of context-dependent phylogenetic models, which capture phenomena such as a strong CpG effect in noncoding regions and a preference for synonymous rather than nonsynonymous substitutions in coding regions; and (2) a novel strategy for incorporating insertions and deletion (indels) into the state-transition structure of the model, which captures the different characteristic patterns of alignment gaps in coding and noncoding regions. We also discuss the technique, previously used in pairwise gene predictors, of explicitly modeling conserved noncoding sequence to help reduce false positive predictions. These methods have been incorporated into an exon prediction program called ExoniPhy, and tested with two large data sets. Experimental results indicate that all three methods produce significant improvements in prediction performance. In combination, they lead to prediction accuracy comparable to that of some of the best available gene predictors, despite several limitations of our current models.

Working drafts are now available for the human, mouse and rat genomes, and other mammalian genome sequences are on the way. We discuss some of the key bioinformatic analysis problems presented by this data, including the problems of assembling the sequence, finding the genes and other functional elements, and reconstructing the evolutionary history of the genomes. Recent comparisons between the human and mouse genomes have revealed that approximately 5% of the human genome appears to be more conserved with the orthologous regions in mouse than can be explained assuming neutral evolution. Is this the portion of the genome under selection for specific functions? How can we use comparative genomics to further pinpoint functional elements? How accurately can we reconstruct the evolutionary history of key parts of the human genome? We briefly outline some recent work (described in more detail in Adam Siepel's talk) combining hidden Markov models, used in bioinformatics to analyse DNA from a single species, with continuous time Markov models of molecular evolution, used to reconstruct evolutionary history of several species. While still a long way from answering these questions, these methods may contribute to such investigations.

We construct several score functions for use in locating unusually conserved regions in a genome-wide search of aligned DNA from two species. We test these functions on regions of the human genome aligned to the mouse genome. These score functions are derived from properties of neutrally evolving sites on the mouse and human genome, and can be adjusted to the local background rate of conservation. The aim of these functions is to try to identify regions of the human genome that are conserved by evolutionary selection, because they have an important function, rather than by chance. We use them to get a very rough estimate of the amount of DNA in the human genome that is under selection.

A few models have appeared in recent years that consider not only the way substitutions occur through evolutionary history at each site of a genome, but also the way the process changes from one site to the next. These models combine phylogenetic models of molecular evolution, which apply to individual sites, and hidden Markov models, which allow for changes from site to site. Besides improving the realism of ordinary phylogenetic models, they are potentially very powerful tools for inference and prediction---for gene finding, for example, or prediction of secondary structure. In this paper, we review progress on combined phylogenetic and hidden Markov models and present some extensions to previous work. Our main result is a simple and efficient method for accommodating higher-order states in the HMM, which allows for context-sensitive models of substitution---that is, models that consider the effects of neighboring bases on the pattern of substitution. We present experimental results indicating that higher-order states, autocorrelated rates, and multiple functional categories all lead to significant improvements in the fit of a combined phylogenetic and hidden Markov model, with the effect of higher-order states being particularly pronounced.

Abstract Motivation: The enormous amount of protein sequence data uncovered by genome research has increased the demand for computer software that can automate the recognition of new proteins. We discuss the relative merits of various automated methods for recognizing G-Protein Coupled Receptors (GPCRs), a superfamily of cell membrane proteins. GPCRs are found in a wide range of organisms and are central to a cellular signalling network that regulates many basic physiological processes. They are the focus of a significant amount of current pharmaceutical research because they play a key role in many diseases. However, their tertiary structures remain largely unsolved. The methods described in this paper use only primary sequence information to make their predictions. We compare a simple nearest neighbor approach (BLAST), methods based on multiple alignments generated by a statistical profile Hidden Markov Model (HMM), and methods, including Support Vector Machines (SVMs), that transform protein sequences into fixed-length feature vectors. Results: The last is the most computationally expensive method, but our experiments show that, for those interested in annotation-quality classification, the results are worth the effort. In two-fold cross-validation experiments testing recognition of GPCR subfamilies that bind a specific ligand (such as a histamine molecule), the errors per sequence at the Minimum Error Point (MEP) were 13.7% for multi-class SVMs, 17.1% for our SVMtree method of hierarchical multi-class SVM classification, 25.5% for BLAST, 30% for profile HMMs, and 49% for classification based on nearest neighbor feature vector Kernel Nearest Neighbor (kernNN). The percentage of true positives recognized before the first false positive was 65% for both SVM methods, 13% for BLAST, 5% for profile HMMs and 4% for kernNN. Availability: We have set up a web server for GPCR subfamily classification based on hierarchical multi-class SVMs at http://www.soe.ucsc.edu/research/compbio/gpcr-subclass. By scanning predicted peptides found in the human genome with the SVMtree server, we have identified a large number of genes that encode GPCRs. A list of our predictions for human GPCRs is available at http://www.soe.ucsc.edu/research/compbio/gpcr˙hg/class˙results. We also provide suggested subfamily classification for 18 sequences previously identified as unclassified Class A (rhodopsin-like) GPCRs in GPCRDB (Horn et al. , Nucleic Acids Res. , 26, 277–281, 1998), available at http://www.soe.ucsc.edu/research/compbio/gpcr/classA˙unclassified/. Contact: [email protected] * To whom correspondence should be addressed.

MOTIVATION: DNA microarray experiments generating thousands of gene expression measurements, are being used to gather information from tissue and cell samples regarding gene expression differences that will be useful in diagnosing disease. We have developed a new method to analyse this kind of data using support vector machines (SVMs). This analysis consists of both classification of the tissue samples, and an exploration of the data for mis-labeled or questionable tissue results. RESULTS: We demonstrate the method in detail on samples consisting of ovarian cancer tissues, normal ovarian tissues, and other normal tissues. The dataset consists of expression experiment results for 97,802 cDNAs for each tissue. As a result of computational analysis, a tissue sample is discovered and confirmed to be wrongly labeled. Upon correction of this mistake and the removal of an outlier, perfect classification of tissues is achieved, but not with high confidence. We identify and analyse a subset of genes from the ovarian dataset whose expression is highly differentiated between the types of tissues. To show robustness of the SVM method, two previously published datasets from other types of tissues or cells are analysed. The results are comparable to those previously obtained. We show that other machine learning methods also perform comparably to the SVM on many of those datasets. AVAILABILITY: The SVM software is available at http://www.cs. columbia.edu/ approximately bgrundy/svm.

We consider adaptive sequential prediction of arbitrary binary sequences when the performance is evaluated using a general loss function. The goal is to predict on each individual sequence nearly as well as the best prediction strategy in a given comparison class of (possibly adaptive) prediction strategies, called experts. By using a general loss function, we generalize previous work on universal prediction, forecasting, and data compression. However, here we restrict ourselves to the case when the comparison class is finite. For a given sequence, we define the regret as the total loss on the entire sequence suffered by the adaptive sequential predictor, minus the total loss suffered by the predictor in the comparison class that performs best on that particular sequence. We show that for a large class of loss functions, the minimax regret is either /spl theta/(log N) or /spl Omega/(/spl radic//spl Lscr/log N), depending on the loss function, where N is the number of predictors in the comparison class and/spl Lscr/ is the length of the sequence to be predicted. The former case was shown previously by Vovk (1990); we give a simplified analysis with an explicit closed form for the constant in the minimax regret formula, and give a probabilistic argument that shows this constant is the best possible. Some weak regularity conditions are imposed on the loss function in obtaining these results. We also extend our analysis to the case of predicting arbitrary sequences that take real values in the interval [0,1].

We discuss some learning problems in molecular biology that arise from the need to interpret the data generated by the Human Genome Project.

We present an improved splice site predictor for the genefinding program Genie. Genie is based on a generalized Hidden Markov Model (GHMM) that describes the grammar of a legal parse of a multi-exon gene in a DNA sequence. In Genie, probabilities are estimated for gene features by using dynamic programming to combine information from multiple content and signal sensors, including sensors that integrate matches to homologous sequences from a database. One of the hardest problems in genefinding is to determine the complete gene structure correctly. The splice site sensors are the key signal sensors that address this problem. We replaced the existing splice site sensors in Genie with two novel neural networks based on dinucleotide frequencies. Using these novel sensors, Genie shows significant improvements in the sensitivity and specificity of gene structure identification. Experimental results in tests using a standard set of annotated genes showed that Genie identified 86% of coding nucleotides correctly with a specificity of 85%, versus 80% and 84% in the older system. In further splice site experiments, we also looked at correlations between splice site scores and intron and exon lengths, as well as at the effect of distance to the nearest splice site on false positive rates.

Learnability in Valiant's PAC learning model has been shown to be strongly related to the existence of uniform laws of large numbers. These laws define a distribution-free convergence property of means to expectations uniformly over classes of random variables. Classes of real-valued functions enjoying such a property are also known as uniform Glivenko-Cantelli classes. In this paper, we prove, through a generalization of Sauer's lemma that may be interesting in its own right, a new characterization of uniform Glivenko-Cantelli classes. Our characterization yields Dudley, Gine´, and Zinn's previous characterization as a corollary. Furthermore, it is the first based on a Gine´, and Zinn's previous characterization as a corollary. Furthermore, it is the first based on a simple combinatorial quantity generalizing the Vapnik-Chervonenkis dimension. We apply this result to obtain the weakest combinatorial condition known to imply PAC learnability in the statistical regression (or “agnostic”) framework. Furthermore, we find a characterization of learnability in the probabilistic concept model, solving an open problem posed by Kearns and Schapire. These results show that the accuracy parameter plays a crucial role in determining the effective complexity of the learner's hypothesis class.

We analyze algorithms that predict a binary value by combining the predictions of several prediction strategies, calledexperts. Our analysis is for worst-case situations, i.e., we make no assumptions about the way the sequence of bits to be predicted is generated. We measure the performance of the algorithm by the difference between the expected number of mistakes it makes on the bit sequence and the expected number of mistakes made by the best expert on this sequence, where the expectation is taken with respect to the randomization in the predictins. We show that the minimum achievable difference is on the order of the square root of the number of mistakes of the best expert, and we give efficient algorithms that achieve this. Our upper and lower bounds have matching leading constants in most cases. We then show how this leads to certain kinds of pattern recognition/learning algorithms with performance bounds that improve on the best results currently know in this context. We also compare our analysis to the case in which log loss is used instead of the expected number of mistakes.

Suppose nature picks a probability measure P/sub /spl theta// on a complete separable metric space X at random from a measurable set P/sub /spl Theta//={P/spl theta/:/spl theta//spl isin//spl Theta/}. Then, without knowing /spl theta/, a statistician picks a measure Q on S. Finally, the statistician suffers a loss D(P/sub 0//spl par/Q), the relative entropy between P/sub /spl theta// and Q. We show that the minimax and maximin values of this game are always equal, and there is always a minimax strategy in the closure of the set of all Bayes strategies. This generalizes previous results of Gallager(1979), and Davisson and Leon-Garcia (1980).

We present a method for condensing the information in multiple alignments of proteins into a mixture of Dirichlet densities over amino acid distributions. Dirichlet mixture densities are designed to be combined with observed amino acid frequencies to form estimates of expected amino acid probabilities at each position in a profile, hidden Markov model or other statistical model. These estimates give a statistical model greater generalization capacity, so that remotely related family members can be more reliably recognized by the model. This paper corrects the previously published formula for estimating these expected probabilities, and contains complete derivations of the Dirichlet mixture formulas, methods for optimizing the mixtures to match particular databases, and suggestions for efficient implementation.

In this paper we introduce and investigate a mathematically rigorous theory of learning curves that is based on ideas from statistical mechanics. The advantage of our theory over the well-established Vapnik-Chervonenkis theory is that our bounds can be considerably tighter in many cases, and are also more reflective of the true behavior (functional form) of learning curves. This behavior can often exhibit dramatic properties such as phase transitions, as well as power law asymptotics not explained by the VC theory. The disadvantages of our theory are that its application requires knowledge of the input distribution, and it is limited so far to finite cardinality function classes. We illustrate our results with many concrete examples of learning curve bounds derived from our theory.

Learnability in Valiant's PAC learning model has been shown to be strongly related to the existence of uniform laws of large numbers. These laws define a distribution-free convergence property of means to expectations uniformly over classes of random variables. Classes of real-valued functions enjoying such a property are also known as uniform Gliveako-Cantelli classes. In this paper we prove, through a generalization of Sauer's lemma that may be interesting in its own right, a new characterization of uniform Glivenko-Cantelli classes. Our characterization yields Dudley, Gine, and Zinn's previous characterization as a corollary. Furthermore, it is the first based on a simple combinatorial quantity generalizing the Vapnik-Chervonenkis dimension. We apply this result to characterize PAC learnability in the statistical regression framework of probabilistic concepts, solving an open problem posed by Kearns and Schapire. Our characterization shows that the accuracy parameter plays a crucial role in determining the effective complexity of the learner's hypothesis class.>

Article How to use expert advice Share on Authors: Nicolò Cesa-Bianchi View Profile , Yoav Freund View Profile , David P. Helmbold View Profile , David Haussler View Profile , Robert E. Schapire View Profile , Manfred K. Warmuth View Profile Authors Info & Claims STOC '93: Proceedings of the twenty-fifth annual ACM symposium on Theory of ComputingJune 1993 Pages 382–391https://doi.org/10.1145/167088.167198Online:01 June 1993Publication History 71citation406DownloadsMetricsTotal Citations71Total Downloads406Last 12 Months8Last 6 weeks3 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

The probably approximately correct (PAC) model of learning from examples is generalized. The problem of learning functions from a set X into a set Y is considered, assuming only that the examples are generated by independent draws according to an unknown probability measure on X*Y. The learner's goal is to find a function in a given hypothesis space of functions from X into Y that on average give Y values that are close to those observed in random examples. The discrepancy is measured by a bounded real-valued loss function. The average loss is called the error of the hypothesis. A theorem on the uniform convergence of empirical error estimates to true error rates is given for certain hypothesis spaces, and it is shown how this implies learnability. A generalized notion of VC dimension that applies to classes of real-valued functions and a notion of capacity for classes of functions that map into a bounded metric space are given. These measures are used to bound the rate of convergence of empirical error estimates to true error rates, giving bounds on the sample size needed for learning using hypotheses in these classes. As an application, a distribution-independent uniform convergence result for certain classes of functions computed by feedforward neural nets is obtained. Distribution-specific uniform convergence results for classes of functions that are uniformly continuous on average are also obtained.>

Valiant's learnability model is extended to learning classes of concepts defined by regions in Euclidean space E n . The methods in this paper lead to a unified treatment of some of Valiant's results, along with previous results on distribution-free convergence of certain pattern recognition algorithms. It is shown that the essential condition for distribution-free learnability is finiteness of the Vapnik-Chervonenkis dimension, a simple combinatorial parameter of the class of concepts to be learned. Using this parameter, the complexity and closure properties of learnable classes are analyzed, and the necessary and sufficient conditions are provided for feasible learnability.

We address the question of when a network can be expected to generalize from m random training examples chosen from some arbitrary probability distribution, assuming that future test examples are drawn from the same distribution. Among our results are the following bounds on appropriate sample vs. network size. Assume 0 < ∊ ≤ 1/8. We show that if m ≥ O(W/∊ log N/∊) random examples can be loaded on a feedforward network of linear threshold functions with N nodes and W weights, so that at least a fraction 1 − ∊/2 of the examples are correctly classified, then one has confidence approaching certainty that the network will correctly classify a fraction 1 − ∊ of future test examples drawn from the same distribution. Conversely, for fully-connected feedforward nets with one hidden layer, any learning algorithm using fewer than Ω(W/∊) random training examples will, for some distributions of examples consistent with an appropriate weight choice, fail at least some fixed fraction of the time to find a weight choice that will correctly classify more than a 1 − ∊ fraction of the future test examples.

The authors consider the problem of predicting (0, 1)-valued functions on R/sup n/ and smaller domains, based on their values on randomly drawn points. Their model is related to L.G. Valiant's learnability model (1984), but does not require the hypotheses used for prediction to be represented in any specified form. The authors first disregard computational complexity and show how to construct prediction strategies that are optimal to within a constant factor for any reasonable class F of target functions. These prediction strategies use the 1-inclusion graph structure from N. Alon et al.'s work on geometric range queries (1987) to minimize the probability of incorrect prediction. They then turn to computationally efficient algorithms. For indicator functions of axis-parallel rectangles and halfspaces in R/sup n/, they demonstrate how their techniques can be applied to construct computational efficient prediction strategies that are optimal to within a constant factor. They compare the general performance of prediction strategies derived by their method to those derived from existing methods in Valiant's learnability theory.>

Article Partitioning and geometric embedding of range spaces of finite Vapnik-Chervonenkis dimension Share on Authors: N. Alon Department of Mathematics, Tel Aviv University, Ramat Aviv, TEL Aviv 69978, Israel Department of Mathematics, Tel Aviv University, Ramat Aviv, TEL Aviv 69978, IsraelView Profile , D. Haussler Computer Science Department, University of California at Santa Cruz, Santa Cruz, CA, USA Computer Science Department, University of California at Santa Cruz, Santa Cruz, CA, USAView Profile , E. Welzl Institutes for Information Processing, Technical University of Graz, Schiesstattgaser 4a, A-8010 GRAZ, Austria Institutes for Information Processing, Technical University of Graz, Schiesstattgaser 4a, A-8010 GRAZ, AustriaView Profile Authors Info & Claims SCG '87: Proceedings of the third annual symposium on Computational geometryOctober 1987 Pages 331–340https://doi.org/10.1145/41958.41994Online:01 October 1987Publication History 22citation296DownloadsMetricsTotal Citations22Total Downloads296Last 12 Months5Last 6 weeks0 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Given a finite set of texts S = { w 1, … , w k } over some fixed finite alphabet Σ, a complete inverted file for S is an abstract data type that provides the functions find ( w ), which returns the longest prefix of w that occurs (as a subword of a word) in S ; freq ( w ), which returns the number of times w occurs in S ; and locations ( w ), which returns the set of positions where w occurs in S . A data structure that implements a complete inverted file for S that occupies linear space and can be built in linear time, using the uniform-cost RAM model, is given. Using this data structure, the time for each of the above query functions is optimal. To accomplish this, techniques from the theory of finite automata and the work on suffix trees are used to build a deterministic finite automaton that recognizes the set of all subwords of the set S . This automaton is then annotated with additional information and compacted to facilitate the desired query functions. The result is a data structure that is smaller and more flexible than the suffix tree.

We present a new technique for half-space and simplex range query using Ο(n) space and Ο(na) query time, where a < d(d-1)/d(d-1) + 1 + γ for all dimensions d ≥ 2 and γ > 0. These bounds are better than those previously published for all d ≥ 2. The technique uses random sampling to build a partition-tree structure. We introduce the concept of an ε-net for an abstract set of ranges to describe the desired result of this random sampling and give necessary and sufficient conditions that a random sample is an ε-net with high probability. We illustrate the application of these ideas to other range query problems.

Article Classifying learnable geometric concepts with the Vapnik-Chervonenkis dimension Share on Authors: A Blumer University of California at Santa Cruz and Department of Mathematics and Computer Science, University of Denver, Denver, Colorado University of California at Santa Cruz and Department of Mathematics and Computer Science, University of Denver, Denver, ColoradoView Profile , A Ehrenfeucht Department of Computer Science, University of Colorado, Boulder, Colorado Department of Computer Science, University of Colorado, Boulder, ColoradoView Profile , D Haussler Department of Mathematics and Computer Science, University of Denver, Denver, Colorado Department of Mathematics and Computer Science, University of Denver, Denver, ColoradoView Profile , M Warmuth Department of Computer and Information Sciences, University of California, Santa Cruz, California Department of Computer and Information Sciences, University of California, Santa Cruz, CaliforniaView Profile Authors Info & Claims STOC '86: Proceedings of the eighteenth annual ACM symposium on Theory of computingNovember 1986 Pages 273–282https://doi.org/10.1145/12130.12158Online:01 November 1986Publication History 83citation790DownloadsMetricsTotal Citations83Total Downloads790Last 12 Months58Last 6 weeks16 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Given a finite set of texts S = {ω1, ..., ωk} over some fixed finite alphabet Σ, a complete inverted file for S is an abstract data type that provides the functions find(ω), which returns the longest prefix of ω which occurs in S; freq(ω), which returns the number of times ω occurs in S; and locations(ω) which returns the set of positions at which ω occurs. We give a data structure to implement a complete inverted file for S which occupies linear space and can be built in linear time, using the uniform cost RAM model. Using this data structure, the time for each of the above query functions is optimal. To accomplish this, we use techniques from the theory of finite automata to build a deterministic finite automaton which recognizes the set of all sub words of the set S. This automaton is then annotated with additional information and compacted to facilitate the desired query functions.