This research aims to utilise ligand-based target prediction to (i) understand the mechanism of action of African natural products (ANPs), (ii) help identify patterns of phylogenetic use in African traditional medicine and (iii) elucidate the mechanism of action of phenotypically active small molecules and natural products with anti-trypanosomal activity. In Chapter 2 the objective was to utilise ligand-based target prediction to understand the mechanism of action of natural products (NPs) from African medicinal plants used against cancer. The Random Forest classifier used in this work compares the similarity of the input compounds from the natural product dataset with compound-target combinations in the training set. The more similar they are in structure, the more likely they are to modulate the same target. Natural products from plants used against cancer in Africa were predicted to modulate targets and pathways directly associated with the disease, thus understanding their mechanism of action e.g. "flap endonuclease 1" and "Mcl-1". The "Keap1-Nrf2 Pathway" and "apoptosis modulation by HSP70", two pathways previously linked to cancer (which are not currently targeted by marketed drugs, but have been of increasing interest in recent years) were predicted to be modulated by ANPs. In Chapter 3, we aimed to identify phylogenetic patterns in medicinal plant use and the role this plays in predicting medicinal activity. We combined chemical, predicted target and phylogenetic information of the natural products to identify patterns of use for plant families containing plant species used against cancer in African, Malay and Indian (Ayurveda) traditional medicine. Plant families that are close phylogenetically were found to produce similar natural products that act on similar targets regardless of their origin. Additionally, phylogenetic patterns were identified for African traditional plant families with medicinal species used against cancer, malaria and human African trypanosomiasis (HAT). We identified plant families that have more medicinal species than would statistically be expected by chance and rationalised this by linking their activity to their unique phyto-chemistry e.g. the napthyl-isoquinoline alkaloids, uniquely produced by Acistrocladaceae and Dioncophyllaceae, are responsible for anti-malarial and anti-trypanosome activity. In Chapter 4, information from target prediction and experimentally validated targets was combined with orthologue data to predict targets of phenotypically active small molecules and natural products screened against Trypanosoma brucei. The predicted targets were prioritised based on their essentiality for the survival of the T. brucei parasite. We predicted orthologues of targets that are essential for the survival of the trypanosome e.g. glycogen synthase kinase 3 (GSK3) and rhodesain. We also identified the biological processes predicted to be perturbed by the compounds e.g. "glycolysis", "cell cycle", "regulation of symbiosis, encompassing mutualism through parasitism" and "modulation of development of symbiont involved in interaction with host". In conclusion, in silico target prediction can be used to predict protein targets of natural products to understand their molecular mechanism of action. Phylogenetic information and phytochemical information of medicinal plants can be integrated to identify plant families with more medicinal species than would be expected by chance.

Traditional medicines (TM) have been used for centuries to treat illnesses, but in many cases their modes-of-action (MOAs) remain unclear. Given the increasing data of chemical ingredients of traditional medicines and the availability of large-scale bioactivity data linking chemical structures to activities against protein targets, we are now in a position to propose computational hypotheses for the MOAs using in silico target prediction. The MOAs were established from supporting literature. The in silico target prediction, which is based on the “Molecular Similarity Principle”, was modelled via two models: a Naïve Bayes Classifier and a Random Forest Classifier. Chapter 2 discovered the relationship of 46 traditional Chinese medicine (TCM) therapeutic action subclasses by mapping them into a dendrogram using the predicted targets. Overall, the most frequent top three enriched targets/pathways were immune-related targets such as tyrosine-protein phosphatase non-receptor type 2 (PTPN2) and digestive system such as mineral absorption. Two major protein families, G-protein coupled receptor (GPCR), and protein kinase family contributed to the diversity of the bioactivity space, while digestive system was consistently annotated pathway motif. Chapter 3 compared the chemical and bioactivity space of 97 anti-cancer plants’ compounds of TCM, Ayurveda and Malay traditional medicine. The comparison of the chemical space revealed that benzene, anthraquinone, flavone, sterol, pentacyclic triterpene and cyclohexene were the most frequent scaffolds in those TM. The annotation of the bioactivity space with target classes showed that kinase class was the most significant target class for all groups. From a phylogenetic tree of the anti-cancer plants, only eight pairs of plants were phylogenetically related at either genus, family or order level. Chapter 4 evaluated synergy score of pairwise compound combination of Shexiang Baoxin Pill (SBP), a TCM formulation for myocardial infarction. The score was measured from the topological properties, pathway dissimilarity and mean distance of all the predicted targets of a combination on a representative network of the disease. The method found four synergistic combinations, ginsenoside Rb3 and cholic acid, ginsenoside Rb2 and ginsenoside Rb3, ginsenoside Rb3 and 11-hydroxyprogesterone and ginsenoside Rb2 and ginsenoside Rd agreed with the experimental results. The modulation of androgen receptor, epidermal growth factor and caspases were proposed for the synergistic actions. Altogether, in silico target prediction was able to discover the bioactivity space of different TMs and elucidate the MOA of multiple formulations and two major health concerns: cancer and myocardial infarction. Hence, understanding the MOA of the traditional medicine could be beneficial in providing testable hypotheses to guide towards finding new molecular entities.

The amount of biomedical literature has been increasing rapidly during the last decade. Text mining techniques can harness this large-scale data, shed light onto complex drug mechanisms, and extract relation information that can support computational polypharmacology. In this work, we introduce CASSANDRA, a fully corpus-based and unsupervised algorithm which uses the MEDLINE indexed titles and abstracts to infer drug gene associations and assist drug repositioning. CASSANDRA measures the Pointwise Mutual Information (PMI) between biomedical terms derived from Gene Ontology (GO) and Medical Subject Headings (MeSH). Based on the PMI scores, drug and gene profiles are generated and candidate drug gene associations are inferred when computing the relatedness of their profiles. Results show that an Area Under the Curve (AUC) of up to 0.88 can be achieved. The algorithm can successfully identify direct drug gene associations with high precision and prioritize them over indirect drug gene associations. Validation shows that the statistically derived profiles from literature perform as good as (and at times better than) the manually curated profiles. In addition, we examine CASSANDRA’s potential towards drug repositioning. For all FDA-approved drugs repositioned over the last 5 years, we generate profiles from publications before 2009 and show that the new indications rank high in these profiles. In summary, co-occurrence based profiles derived from the biomedical literature can accurately predict drug gene associations and provide insights onto potential repositioning cases.

Small nucleolar RNAs (snoRNAs) are one of the most abundant and evolutionary ancient group of small non-coding RNAs. Their main function is to target chemical modifications of ribosomal RNAs (rRNAs) and small nuclear (snRNAs). They fall into two classes, box C/D snoRNAs and box H/ACA snoRNAs, which are clearly distinguished by conserved sequence motifs and the type of modification that they govern. The box H/ACA snoRNAs are responsible for targeting pseudouridylation sites and the box C/D snoRNAs for directing 2’-O-methylation of ribonucleotides. A subclass that localize to the Cajal bodies, termed scaRNAs, are responsible for methylation and pseudouridylation of snRNAs. In addition an amazing diversity of non-canonical functions of individual snoRNAs arose. The modification patterns in rRNAs and snRNAs are retained during evolution making it even possible to project them from yeast onto human. The stringent conservation of modification sites and the slow evolution of rRNAs and snRNAs contradicts the rapid evolution of snoRNA sequences. Recent studies that incorporate high-throughput sequencing experiments still identify undetected snoRNAs even in well studied organisms as human. The snoRNAbase, which has been the standard database for human snoRNAs has not been updated ince 2006 and misses these new data. Along with the lack of a centralized data collection across species, which incorporates also snoRNA class specific characteristics the need to integrate distributed data from literature and databases into a comprehensive snoRNA set arose. Although several snoRNA studies included pro forma target predictions in individual species and more and more studies focus on non-canonical functions of subclasses a systematic survey on the guiding function and especially functional homologies of snoRNAs was not available. To establish a sound set of snoRNAs a computational snoRNA annotation pipeline, named snoStrip that identifies homologous snoRNAs in related species was employed. For large scale investigation of the snoRNA function, state-of-the-art target pedictions were performed with our software RNAsnoop and PLEXY. Further, a new measure the Interaction Conservation Index (ICI) was developed to evaluate the conservation of snoRNA function. The snoStrip pipeline was applied to vertebrate species, where the genome sequence has been available. In addition, it was used in several ncRNA annotation studies (48 avian, spotted gar) of newly assembled genomes to contribute the snoRNA genes. Detailed target analysis of the new vertebrate snoRNA set revealed that in general functions of homologous snoRNAs are evolutionarily stable, thus, members of the same snoRNA family guide equivalent modifications. The conservation of snoRNA sequences is high at target binding regions while the remaining sequence varies significantly. In addition to elucidating principles of correlated evolution it was possible, with the help of the ICI measure, to assign functions to previously orphan snoRNAs and to associate snoRNAs as partners to known but so far unexplained chemical modifications. As further pattern redundant guiding became apparent. For many modification sites more than one snoRNA encodes the appropriate antisense element (ASE), which could ensure constant modification through snoRNAs that have different expression patterns. Furthermore, predictions of snoRNA functions in conjunction with sequence conservation could identify distant homologies. Due to the high overall entropy of snoRNA sequences, such relationships are hard to detect by means of sequence homology search methods alone. The snoRNA interaction network was further expanded through novel snoRNAs that were detected in data from high-throughput experiments in human and mouse. Through subsequent target analysis the new snoRNAs could immediately explain known modifications that had no appropriate snoRNA guide assigned before. In a further study a full catalog of expressed snoRNAs in human was provided. Beside canonical snoRNAs also recent findings like AluACAs, sno-lncRNAs and extraordinary short SNORD-like transcripts were taken into account. Again the target analysis workflow identified undetected connections between snoRNA guides and modifications. Especially some species/clade specific interactions of SNORD-like genes emerged that seem to act as bona fide snoRNA guides for rRNA and snRNA modifications. For all high confident new snoRNA genes identified during this work official gene names were requested from the HUGO Gene Nomenclature Committee (HGNC) avoiding further naming confusion.

RNA is a highly versatile molecule with functions that span from being a messenger in the transfer from DNA to protein, a catalytic molecule important for key processes in the cell to a regulator of gene expression. The post-genomic era and the use of new techniques to sequence RNAs have dramatically increased the number of regulatory RNAs during the last decade. Many of these are antisense RNAs, as for example the miRNA in eukaryotes and most sRNAs in bacteria. Antisense RNAs bind to specific targets by basepairing and thereby regulate their expression. A major step towards an understanding of the biological role of a miRNA or an sRNA is taken when one identifies which target it regulates. We have used RNA libraries to study the RNA interference pathway during development in the unicellular model organism Dictyostelium discoideum. We have also, by combining computational and experimental methods, discovered the first miRNAs in this organism and shown that they have different expression profiles during development. In parallel, we have developed a novel approach to predict targets for sRNAs in bacteria and used it to discover sRNA/target RNA interactions in the model organism Escherichia coli. We have found evidence for, and further characterized, three of these predicted sRNA/target interactions. For instance, the sRNA MicA is important for regulation of the outer membrane protein OmpA, the sRNAs OmrA and OmrB regulate the transcription factor CsgD, which is important in the sessile lifestyle of E. coli, and MicF regulates its own expression in a feed forward loop via the regulatory protein Lrp. In conclusion, we have discovered novel antisense RNAs, e.g. miRNAs in D. discoideum, developed an approach to identify targets for antisense RNAs, i.e. a target prediction program for sRNAs in bacteria, and verified and characterized some of the predicted antisense RNA interactions.

<p>MicroRNAs (miRNAs) are small RNAs that have important roles in post-transcriptional gene regulation in a wide range of species. This regulation is controlled by having miRNAs directly bind to a target messenger RNA (mRNA), causing it to be destabilized and degraded, or translationally repressed. Identifying miRNA targets has been a large area of focus for study; however, a lack of generally high-throughput experiments to validate direct miRNA targeting has been a limiting factor. To overcome these limitations, computational methods have become crucial for understanding and predicting miRNA-gene target interactions.</p><p>While a variety of computational tools exist for predicting miRNA targets, many of them are focused on a similar feature set for their prediction. These commonly used features are complementarity to 5'seed of miRNAs and evolutionary conservation. Unfortunately, not all miRNA target sites are conserved or adhere to canonical seed complementarity. Seeking to address these limitations, several studies have included energy features of mRNA:miRNA duplex formation as alternative features. However, different independent evaluations reported conflicting results on the reliability of energy-based predictions. Here, we reassess the usefulness of energy features for mammalian target prediction, aiming to relax or eliminate the need for perfect seed matches and conservation requirement.</p><p>We detect significant differences of energy features at experimentally supported human miRNA target sites and at genome-wide interaction sites to Argonaute (AGO) protein family members, which are essential parts of the miRNA machinery complex. This trend is confirmed on data sets that assay the effect of miRNAs on mRNA and protein expression changes, where a statistically significant change in expression is noted when compared to the control. Furthermore, our method also allows for prediction of strictly imperfect sites, as well as non-conserved targets.</p><p>Recently, new methods for identifying direct miRNA binding have been developed, which provides us with additional sources of information for miRNA target prediction. While some computational target predictions tools have begun to incorporate this information, they still rely on the presence of a seed match in the AGO-bound windows without accounting for the possibility of variations. </p><p>We investigate the usefulness of the site level direct binding evidence in miRNA target identification and propose a model that incorporates multiple different features along with the AGO-interaction data. Our method outperforms both an ad hoc strategy of seed match searches as well as an existing target prediction tool, while still allowing for predictions of sites other than a long perfect seed match. Additionally, we show supporting evidence for a class of non-canonical sites as bound targets. Our model can be extended to predict additional types of imperfect sites, and can also be readily modified to include additional features that may produce additional improvements.</p> / Dissertation

La sélection d'entités est une des tâches courantes et fondamentales de l'interaction. En environnement de réalité virtuelle, cette tâche s'effectue dans les trois dimensions, mais s'accompagne de difficultés inhérentes à ces environnements, comme une mauvaise perception de la profondeur. Des techniques existent pour pallier ces obstacles. Nous présentons trois nouvelles méthodes améliorant les performances de l'utilisateur durant les différentes étapes du processus de sélection. Le principe du Ring Concept permet de localiser des objets non visibles dans la scène. La technique Starfish guide le mouvement de l'utilisateur vers sa cible. Enfin, l'algorithme SPEED prédit le point d'arrivée d'un mouvement de sélection. / Selection is one of the most current and fundamental interaction tasks. In a virtual reality environment, this task is performed in the three dimensions, but is accompanied by difficulties inherent in these environments, such as poor depth perception. Several techniques exist to overcome these obstacles. We present three new methods for improving the performance of the user during the various phases of the selection process. The Ring Concept principle can locate non-visible objects in the scene. The Starfish technique guides the movement of the user to the target. Finally, the SPEED algorithm predicts the endpoint of a selection movement.

Transcription Activator-Like Effectors (TALEs) can be used as adaptable DNAbinding modules to create site-specific chimeric nucleases or synthetic transcriptional regulators. The central repeat domain mediates specific DNA binding via hypervariable repeat di-residues (RVDs). This DNA-Binding Domain can be engineered to bind preferentially to any user-selected DNA sequence if engineered appropriately. Therefore, TALEs and their derivatives have become indispensable molecular tools in site-specific manipulation of genes and genomes. This thesis revolves around two problems: in silico design and improved binding site prediction of TALEs. In the first part, a study is shown where TALEs are successfully designed in silico and validated in laboratory to yield the anticipated effects on selected genes. Software is developed to accompany the process of designing and prediction of binding sites. I expanded the functionality of the software to be used as a more generic set of tools for the design, target and offtarget searching. Part two contributes a method and associated toolkit developed to allow users to design in silico optimized synthetic TALEs with user-defined specificities for various experimental purposes. This method is based on a mutual relationship of three consecutive tandem repeats in the DNA-binding domain. This approach revealed positional and compositional bias behind the binding of TALEs to DNA. In conclusion, I developed methods, approaches, and software to enhance the functionality of synthetic TALEs, which should improve understanding of TALEs biology and will further advance genome-engineering applications in various organisms and cell types.

Small nucleolar RNAs (snoRNAs) are one of the most abundant and evolutionary ancient group of small non-coding RNAs. Their main function is to target chemical modifications of ribosomal RNAs (rRNAs) and small nuclear (snRNAs). They fall into two classes, box C/D snoRNAs and box H/ACA snoRNAs, which are clearly distinguished by conserved sequence motifs and the type of modification that they govern. The box H/ACA snoRNAs are responsible for targeting pseudouridylation sites and the box C/D snoRNAs for directing 2’-O-methylation of ribonucleotides. A subclass that localize to the Cajal bodies, termed scaRNAs, are responsible for methylation and pseudouridylation of snRNAs. In addition an amazing diversity of non-canonical functions of individual snoRNAs arose. The modification patterns in rRNAs and snRNAs are retained during evolution making it even possible to project them from yeast onto human. The stringent conservation of modification sites and the slow evolution of rRNAs and snRNAs contradicts the rapid evolution of snoRNA sequences. Recent studies that incorporate high-throughput sequencing experiments still identify undetected snoRNAs even in well studied organisms as human. The snoRNAbase, which has been the standard database for human snoRNAs has not been updated ince 2006 and misses these new data. Along with the lack of a centralized data collection across species, which incorporates also snoRNA class specific characteristics the need to integrate distributed data from literature and databases into a comprehensive snoRNA set arose. Although several snoRNA studies included pro forma target predictions in individual species and more and more studies focus on non-canonical functions of subclasses a systematic survey on the guiding function and especially functional homologies of snoRNAs was not available. To establish a sound set of snoRNAs a computational snoRNA annotation pipeline, named snoStrip that identifies homologous snoRNAs in related species was employed. For large scale investigation of the snoRNA function, state-of-the-art target pedictions were performed with our software RNAsnoop and PLEXY. Further, a new measure the Interaction Conservation Index (ICI) was developed to evaluate the conservation of snoRNA function. The snoStrip pipeline was applied to vertebrate species, where the genome sequence has been available. In addition, it was used in several ncRNA annotation studies (48 avian, spotted gar) of newly assembled genomes to contribute the snoRNA genes. Detailed target analysis of the new vertebrate snoRNA set revealed that in general functions of homologous snoRNAs are evolutionarily stable, thus, members of the same snoRNA family guide equivalent modifications. The conservation of snoRNA sequences is high at target binding regions while the remaining sequence varies significantly. In addition to elucidating principles of correlated evolution it was possible, with the help of the ICI measure, to assign functions to previously orphan snoRNAs and to associate snoRNAs as partners to known but so far unexplained chemical modifications. As further pattern redundant guiding became apparent. For many modification sites more than one snoRNA encodes the appropriate antisense element (ASE), which could ensure constant modification through snoRNAs that have different expression patterns. Furthermore, predictions of snoRNA functions in conjunction with sequence conservation could identify distant homologies. Due to the high overall entropy of snoRNA sequences, such relationships are hard to detect by means of sequence homology search methods alone. The snoRNA interaction network was further expanded through novel snoRNAs that were detected in data from high-throughput experiments in human and mouse. Through subsequent target analysis the new snoRNAs could immediately explain known modifications that had no appropriate snoRNA guide assigned before. In a further study a full catalog of expressed snoRNAs in human was provided. Beside canonical snoRNAs also recent findings like AluACAs, sno-lncRNAs and extraordinary short SNORD-like transcripts were taken into account. Again the target analysis workflow identified undetected connections between snoRNA guides and modifications. Especially some species/clade specific interactions of SNORD-like genes emerged that seem to act as bona fide snoRNA guides for rRNA and snRNA modifications. For all high confident new snoRNA genes identified during this work official gene names were requested from the HUGO Gene Nomenclature Committee (HGNC) avoiding further naming confusion.

The amount of biomedical literature has been increasing rapidly during the last decade. Text mining techniques can harness this large-scale data, shed light onto complex drug mechanisms, and extract relation information that can support computational polypharmacology. In this work, we introduce CASSANDRA, a fully corpus-based and unsupervised algorithm which uses the MEDLINE indexed titles and abstracts to infer drug gene associations and assist drug repositioning. CASSANDRA measures the Pointwise Mutual Information (PMI) between biomedical terms derived from Gene Ontology (GO) and Medical Subject Headings (MeSH). Based on the PMI scores, drug and gene profiles are generated and candidate drug gene associations are inferred when computing the relatedness of their profiles. Results show that an Area Under the Curve (AUC) of up to 0.88 can be achieved. The algorithm can successfully identify direct drug gene associations with high precision and prioritize them over indirect drug gene associations. Validation shows that the statistically derived profiles from literature perform as good as (and at times better than) the manually curated profiles. In addition, we examine CASSANDRA’s potential towards drug repositioning. For all FDA-approved drugs repositioned over the last 5 years, we generate profiles from publications before 2009 and show that the new indications rank high in these profiles. In summary, co-occurrence based profiles derived from the biomedical literature can accurately predict drug gene associations and provide insights onto potential repositioning cases.