The human genome holds an extraordinary trove of information about human development, physiology, medicine and evolution. Here we report the results of an international collaboration to produce and make freely available a draft sequence of the human genome. We also present an initial analysis of the data, describing some of the insights that can be gleaned from the sequence.

/pdf/initial-sequencing-and-analysis-of-the-human-genome-5dapk1rsx1.pdf

Initial sequencing and analysis of the human genome.

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

We have developed a new set of algorithms, collectively called "Velvet," to manipulate de Bruijn graphs for genomic sequence assembly. A de Bruijn graph is a compact representation based on short words (k-mers) that is ideal for high coverage, very short read (25-50 bp) data sets. Applying Velvet to very short reads and paired-ends information only, one can produce contigs of significant length, up to 50-kb N50 length in simulations of prokaryotic data and 3-kb N50 on simulated mammalian BACs. When applied to real Solexa data sets without read pairs, Velvet generated contigs of approximately 8 kb in a prokaryote and 2 kb in a mammalian BAC, in close agreement with our simulated results without read-pair information. Velvet represents a new approach to assembly that can leverage very short reads in combination with read pairs to produce useful assemblies.

https://www.cs.utexas.edu/~tandy/velvet.pdf

Velvet: Algorithms for de novo short read assembly using de Bruijn graphs

We have catalogued the protein kinase complement of the human genome (the "kinome") using public and proprietary genomic, complementary DNA, and expressed sequence tag (EST) sequences. This provides a starting point for comprehensive analysis of protein phosphorylation in normal and disease states, as well as a detailed view of the current state of human genome analysis through a focus on one large gene family. We identify 518 putative protein kinase genes, of which 71 have not previously been reported or described as kinases, and we extend or correct the protein sequences of 56 more kinases. New genes include members of well-studied families as well as previously unidentified families, some of which are conserved in model organisms. Classification and comparison with model organism kinomes identified orthologous groups and highlighted expansions specific to human and other lineages. We also identified 106 protein kinase pseudogenes. Chromosomal mapping revealed several small clusters of kinase genes and revealed that 244 kinases map to disease loci or cancer amplicons.

http://science.sciencemag.org/content/sci/298/5600/1912.full.pdf

The Protein Kinase Complement of the Human Genome

We will review some of the major results in random graphs and some of the more challenging open problems. We will cover algorithmic and structural questions. We will touch on newer models, including those related to the WWW.

Random graphs

Gene duplicates have been shown to evolve at different rates. Here we further investigate the mechanism and functional underpinning of this phenomenon by assessing asymmetric evolution specifically within functional domains of gene duplicates. Based on duplicate genes in five teleost fishes resulting from a whole genome duplication event, we first show that a Fisher Exact test based approach to detect asymmetry is more sensitive than the previously used Likelihood Ratio test. Using our Fisher Exact test, we found that the evolutionary rate asymmetry in the overall protein is largely explained by the asymmetric evolution within specific protein domains. Moreover, among cases of asymmetrically evolving domains, for the gene copy containing a fast evolving domain, the non-synonymous substitutions often cluster within the fast evolving domain. We found that rare substitutions were preferred within asymmetrically evolving domains suggestive of functional divergence. While overall ~32 % of the domains tested were found to be evolving asymmetrically, certain protein domains such as the Tyrosine and Ser/Thr Kinase domains had a much greater prevalence of asymmetric evolution. Finally, based on the spatial expression of Zebra fish duplicate proteins during development, we found that protein pairs containing asymmetrically evolving domains had a greater divergence in gene expression as compared to the duplicate proteins that did not exhibit asymmetric evolution. Taken together, our results suggest that the previously observed asymmetry in the overall duplicate protein evolution is largely due to divergence of specific domains of the protein, and coincides with divergence in spatial expression domains.

/pdf/functional-divergence-of-gene-duplicates-a-domain-centric-41mocif83a.pdf

Functional divergence of gene duplicates – a domain-centric view

Motivation: In a tri-partite biological network of transcription factors, their putative target genes, and the tissues in which the target genes are differentially expressed, a tightly inter-connected (dense) subgraph may reveal knowledge about tissue specific transcription regulation mediated by a specific set of transcription factors—a tissue-specific transcriptional module. This is just one context in which an efficient computation of dense subgraphs in a multi-partite graph is needed. Result: Here we report a generic stochastic search based method to compute dense subgraphs in a graph with an arbitrary number of partitions and an arbitrary connectivity among the partitions. We then use the tool to explore tissue-specific transcriptional regulation in the human genome. We validate our findings in Skeletal muscle based on literature. We could accurately deduce biological processes for transcription factors via the tri-partite clusters of transcription factors, genes, and the functional annotation of genes. Additionally, we propose a few previously unknown TF-pathway associations and tissue-specific roles for certain pathways. Finally, our combined analysis of Cardiac, Skeletal, and Smooth muscle data recapitulates the evolutionary relationship among the three tissues. Contact: sridharh@pcbi.upenn.edu

/pdf/dense-subgraph-computation-via-stochastic-search-58nndwk18z.pdf

Dense subgraph computation via stochastic search

The molecular mechanisms that regulate and coordinate signaling between the extracellular matrix (ECM) and cells contributing to the developing vasculature are complex and poorly understood. Myocardin-like protein 2 (MKL2) is a transcriptional co-activator that in response to RhoA and cytoskeletal actin signals physically associates with serum response factor (SRF), activating a subset of SRF-regulated genes. We now report the discovery of a previously undescribed MKL2/TGFβ signaling pathway in embryonic stem (ES) cells that is required for maturation and stabilization of the embryonic vasculature. Mkl2–/– null embryos exhibit profound derangements in the tunica media of select arteries and arterial beds, which leads to aneurysmal dilation, dissection and hemorrhage. Remarkably, TGFβ expression, TGFβ signaling and TGFβ-regulated genes encoding ECM are downregulated in Mkl2–/– ES cells and the vasculature of Mkl2–/– embryos. The gene encoding TGFβ2, the predominant TGFβ isoform expressed in vascular smooth muscle cells and embryonic vasculature, is activated directly via binding of an MKL2/SRF protein complex to a conserved CArG box in the TGFβ2 promoter. Moreover, Mkl2–/– ES cells exhibit derangements in cytoskeletal organization, cell adhesion and expression of ECM that are rescued by forced expression of TGFβ2. Taken together, these data demonstrate that MKL2 regulates a conserved TGF-β signaling pathway that is required for angiogenesis and ultimately embryonic survival.

/pdf/myocardin-like-protein-2-regulates-tgfb-signaling-in-3aayefbzpr.pdf

Myocardin-like protein 2 regulates TGFβ signaling in embryonic stem cells and the developing vasculature

Polyadenylation (poly(A)) of mRNA plays a critical role in regulating gene expression. Identifying the sequence, structural, and epigenomic determinants of poly(A) site usage is an important long t...

https://www.tandfonline.com/doi/pdf/10.4161/rna.8.3.15194

Epigenomic and RNA structural correlates of polyadenylation.

Sequencing by hybridization (SBH) is a promising alternative to the classical DNA sequencing approaches. However, the resolving power of SBH is rather low: with 64kb sequencing chips, unknown DNA fragments only as long as 200 bp can be reconstructed in a single SBH experiment. To improve the resolving power of SBH, positional SBH (PSBH) has recently been suggested; this allows (with additional experimental work) approximate positions of every l-tuple in a target DNA fragment to be measured. We study the positional Eulerian path problem motivated by PSBH. The input to the positional eulerian path problem is an Eulerian graph G(V, E) in which every edge has an associated range of integers and the problem is to find an Eulerian path e1,...,e/E/ in G such that the range of ei contains i. We show that the positional Eulerian path problem is NP-complete even when the maximum out-degree (in-degree) of any vertex in the graph is 2. On a positive note we present polynomial algorithms to solve a special case of PSBH (bounded PSBH), where the range of the allowed positions for any edge is bounded by a constant (it corresponds to accurate experimental measurements of positions in PSBH). Moreover, if the positions of every l-tuple in an unknown DNA fragment of length n are measured with O(log n) error, then our algorithm runs in polynomial time. We also present an estimate of the resolving power of PSBH for a more realistic case when positions are measured with theta (n) error.

/pdf/positional-sequencing-by-hybridization-363apuvuqx.pdf

Sridhar Hannenhalli

Papers

Functional divergence of gene duplicates – a domain-centric view

Dense subgraph computation via stochastic search

Myocardin-like protein 2 regulates TGFβ signaling in embryonic stem cells and the developing vasculature

Epigenomic and RNA structural correlates of polyadenylation.

Positional sequencing by hybridization