References of "Hiard, Samuel"
     in
Bookmark and Share    
Full Text
See detailTrimming the complexity of Ranking by Pairwise Comparison
Hiard, Samuel ULg

Doctoral thesis (2013)

In computer science research, and more specifically in bioinformatics, the size of databases never stops to increase. This can be an issue when trying to answer questions that imply algorithms in ... [more ▼]

In computer science research, and more specifically in bioinformatics, the size of databases never stops to increase. This can be an issue when trying to answer questions that imply algorithms in nonlinear polynomial time with regards to the number of objects in the database, the number of attributes or the number of associated labels per objects. This is the case of the Ranking by Pairwise Comparison (RPC) algorithm. This algorithm builds a model which is able to predict the label preference for a given object, but the computation needs to be performed in an order of N*(N-1)/2 in terms of the number N of labels. Indeed, a pairwise comparator model is needed for each possible pair of labels. Our hypothesis is that a significant part of the set of comparators often contains redundancy and/or noise, so that trimming the set could be beneficiary. We implemented several methods, starting from the simplest one, which merely chooses a set of T comparators (T < N*(N-1)/2) at random, to a more complex approach based on partially randomized greedy search. This thesis will provide a detailed overview of the context we are working in, provide the reader with required background, describe existing preference learning algorithms including RPC, investigate on possible trimming methods and their accuracy, then will conclude on the relevance and robustness of the trimming approximation. After implementing and executing the procedure, we could see that using between N/2 and 2N comparators was sufficient to keep up with the original RPC algorithm, as long as a smart trimming method is used, and sometimes even outperforms it on noisy datasets. Also, comparing the use of base models in regression mode vs. classification mode showed that models built in regression mode may be more robust when using the original RPC. We thus empirically show that, in the particular case of RPC, reducing the complexity of the method gives similar or better results, which means that problems that could not be addressed by this algorithm, or at least not in an acceptable period of time, now can be. We also found that the regression mode yields RPC to be often more robust regarding its base learner parameters, meaning that the quest of optimality, which can also be time-consuming, is less difficult. Yet research on this topic is not over, and we could think of different means to further improve the RPC algorithm or investigate other innovative approaches, which will be discussed in the future work section. Also, the trimming method is not limited to RPC and could be applied to other algorithms which aggregate information provided by a set of models, e.g. the whole multitude of ensemble models used in machine learning. [less ▲]

Detailed reference viewed: 78 (24 ULg)
Full Text
Peer Reviewed
See detailComparator selection for RPC with many labels
Hiard, Samuel ULg; Geurts, Pierre ULg; Wehenkel, Louis ULg

in ECAI 2012 : 20th European Conference on Artificial Intelligence : 27-31 August 2012, Montpellier, France (2012, August)

The Ranking by Pairwise Comparison algorithm (RPC) is a well established label ranking method. However, its complexity is of O(N²) in the number N of labels. We present algorithms for selection, before ... [more ▼]

The Ranking by Pairwise Comparison algorithm (RPC) is a well established label ranking method. However, its complexity is of O(N²) in the number N of labels. We present algorithms for selection, before model construction, a subset of comparators of size O(N), to reduce the computational complexity without loss in accuracy. [less ▲]

Detailed reference viewed: 70 (20 ULg)
Full Text
See detailUsing Class-probability Models instead of Hard Classifiers as Base Learners in the Ranking by Pairwise Comparison Algorithm
Hiard, Samuel ULg; Wehenkel, Louis ULg

in Thatcher, Steve (Ed.) ICMLC 2011 3rd International Conference on Machine Learning and Computing Volume 1 (2011, February)

In the field of Preference Learning, the Ranking by Pairwise Comparison algorithm (RPC) consists of using the learning sample to derive pairwise comparators for each possible pair of class labels, and ... [more ▼]

In the field of Preference Learning, the Ranking by Pairwise Comparison algorithm (RPC) consists of using the learning sample to derive pairwise comparators for each possible pair of class labels, and then aggregating the predictions of the whole set of pairwise comparators for a given object in order to produce a global ranking of the class labels. In its standard form, RPC uses hard binary classifiers assigning an integer (0/1) score to each class concerned by a pairwise comparison. In the present work, we compare this setting with a modified version of RPC, where soft binary class-probability models replace the binary classifiers. To this end, we compare ensembles of extremely randomized classprobability estimation trees with ensembles of extremely randomized classification trees. We empirically show that both approaches lead to equivalent results in terms of Spearman’s rho value when using the optimal settings of their metaparameters. However, we also show that in the context of small and noisy datasets (e.g. with partial ranking information) the use of class-probability models is more robust with respect to variations of its meta-parameter values than the hard classifier ensembles. This suggests that using (soft) class-probability comparators is a sensible option in the context of RPC approaches. [less ▲]

Detailed reference viewed: 72 (17 ULg)
Full Text
Peer Reviewed
See detailPatrocles: a database of polymorphic miRNA-mediated gene regulation in vertebrates
Hiard, Samuel ULg; Charlier, Carole ULg; Coppieters, Wouter ULg et al

in Nucleic Acids Research (2010), 38(Database), 640-651

The Patrocles database (http://www.patrocles.org/) compiles DNA sequence polymorphisms (DSPs) that are predicted to perturb miRNA-mediated gene regulation. Distinctive features include: (i) the coverage ... [more ▼]

The Patrocles database (http://www.patrocles.org/) compiles DNA sequence polymorphisms (DSPs) that are predicted to perturb miRNA-mediated gene regulation. Distinctive features include: (i) the coverage of seven vertebrate species in its present release, aiming for more when information becomes available, (ii) the coverage of the three compartments involved in the silencing process (i.e. targets, miRNA precursors and silencing machinery), (iii) contextual information that enables users to prioritize candidate ‘Patrocles DSPs’, including graphical information on miRNA-target coexpression and eQTL effect of genotype on target expression levels, (iv) the inclusion of Copy Number Variants and eQTL information that affect miRNA precursors as well as genes encoding components of the silencing machinery and (v) a tool (Patrocles finder) that allows the user to determine whether her favorite DSP may perturb miRNA-mediated gene regulation of custom target sequences. To support the biological relevance of Patrocles' content, we searched for signatures of selection acting on ‘Patrocles single nucleotide polymorphisms (pSNPs)’ in human and mice. As expected, we found a strong signature of purifying selection against not only SNPs that destroy conserved target sites but also against SNPs that create novel, illegitimate target sites, which is reminiscent of the Texel mutation in sheep. [less ▲]

Detailed reference viewed: 164 (98 ULg)
Full Text
See detailPatrocles: a database of polymorphic miR-mediated gene regulation in vertebrates
Baurain, Denis ULg; Hiard, Samuel ULg; Coppieters, Wouter ULg et al

Scientific conference (2009, September 29)

Detailed reference viewed: 15 (4 ULg)
Full Text
See detailPatrocles: a database of polymorphic miRNA-mediated gene regulation
Hiard, Samuel ULg; Baurain, Denis ULg; Coppieters, Wouter ULg et al

Conference (2008, March 03)

Detailed reference viewed: 43 (3 ULg)
Full Text
See detailPREDetector : Prokaryotic Regulatory Element Detector
Hiard, Samuel ULg; Rigali, Sébastien ULg; Colson, Séverine ULg et al

Poster (2007, November 12)

Background: In the post-genomic area, in silico predictions of regulatory networks are considered as a powerful approach to decipher and understand biological pathways within prokaryotic cells. The ... [more ▼]

Background: In the post-genomic area, in silico predictions of regulatory networks are considered as a powerful approach to decipher and understand biological pathways within prokaryotic cells. The emergence of position weight matrices based programs has facilitated the access to this approach. However, a tool that automatically estimates the reliability of the predictions and would allow users to extend predictions in genomic regions generally regarded with no regulatory functions was still highly demanded. Result: Here, we introduce PREDetector, a tool developed for predicting regulons of DNA-binding proteins in prokaryotic genomes that (i) automatically predicts, scores and positions potential binding sites and their respective target genes, (ii) includes the downstream co-regulated genes, (iii) extends the predictions to coding sequences and terminator regions, (iv) saves private matrices and allows predictions in other genomes, and (v) provides an easy way to estimate the reliability of the predictions. Conclusion: We present, with PREDetector, an accurate prokaryotic regulon prediction tool that maximally answers biologists’ requests. PREDetector can be downloaded freely at http://www.montefiore.ulg.ac.be/~hiard/predetectorfr.html [less ▲]

Detailed reference viewed: 27 (1 ULg)
Full Text
See detailCompiling polymorphic miRNA-target interactions: the Patrocles database.
Hiard, Samuel ULg; Tordoir, Xavier ULg; Coppieters, Wouter ULg et al

Poster (2007, November 12)

Using positional cloning, we have recently identified the mutation responsible for muscular phenotype of the Texel sheep. It is located in the 3’UTR of the GDF8 gene - a known developmental repressor of ... [more ▼]

Using positional cloning, we have recently identified the mutation responsible for muscular phenotype of the Texel sheep. It is located in the 3’UTR of the GDF8 gene - a known developmental repressor of muscle growth - and creates an illegitimate target site for miRNA expressed in the same tissue. This causes miRNA-mediated translation inhibition of mutant GDF8 transcripts which leads to muscle hypertrophy. We followed up on this finding by searching for common polymorphisms and mutations that affect either (i) RNAi silencing machinery components, (ii) miRNA precursors or (iii) target sites. These might likewise alter miRNA-target interaction and could be responsible for substantial differences in gene expression level. They have been compiled in a public database (“Patrocles”: www.patrocles.org), where they are classified in (i) DNA sequence polymorphisms (DSP) affecting the silencing machinery, (ii) DSP affecting miRNA structure or expression and (iii) DSP affecting miRNA target sites. DSP from the last category were organized in four classes: destroying a target site conserved between mammals (DC), destroying a non-conserved target site (DNC), creating a non-conserved target site (CNC), or shifting a target site (S). To aid in the identification of the most relevant DSP (such as those were a target site is created in an antitarget gene), we have quantified the level of coexpression for all miRNA-gene pairs. Analysis of the numbers of Patrocles-DSP as well as their allelic frequency distribution indicates that a substantial proportion of them undergo purifying selection. The signature of selection was most pronounced for the DC class but was significant for the DNC and CNC class as well, suggesting that a significant proportion of non-conserved targets is truly functional. The Patrocles database allowed for the selection of DSP that are likely to affect gene function and possibly disease susceptibility. The effect of these DSP is being studied both in vitro and in vivo. In conclusion, Patrocles-DSP could be widespread and underlie an appreciable amount of phenotypic variation, including common disease susceptibility. [less ▲]

Detailed reference viewed: 24 (1 ULg)
Full Text
See detailDetection of micro-RNA/gene interactions involved in angiogenesis using machine learning techniques
Huynh-Thu, Vân Anh ULg; Hiard, Samuel ULg; Geurts, Pierre ULg et al

Poster (2007, September)

Motivation: Angiogenesis is the process responsible for the growth of new blood vessels from existing ones. It is also associated with the development of cancer, as tumors need to be irrigated by blood ... [more ▼]

Motivation: Angiogenesis is the process responsible for the growth of new blood vessels from existing ones. It is also associated with the development of cancer, as tumors need to be irrigated by blood vessels for growing. New cancer therapies appear that exploit angiogenesis inhibitors, also called angiostatic agents, to asphyxiate and starve the tumors. Better understanding the regulatory mechanisms that control angiogenesis is thus fundamental. Recently, short non-coding RNA molecules, called micro-RNAs, have been discovered that are involved in post- transcriptional regulation of gene expressions. These molecules bind to RNA messengers following the base pairing rules, preventing them from being translated into proteins and/or tagging them for degradation. The main goal of this work is to use computational approaches to identify micro-RNAs involved in angiogenesis. Method: In order to identify genes involved in angiogenesis, bovine endothelial cells were treated by a known angiogenesis inhibitor [1], prolactin 16K, and their gene expression profile was compared to the profile of untreated cells. The genes were then divided into three classes: up-regulated, down-regulated, and unaffected genes. The 3'UTR regions of these genes were then analysed by machine learning techniques. Different approaches were considered. First, we described each gene by a vector of motif counts in their 3'UTR regions and used machine learning techniques to rank the motifs according to their relevance for separating the genes into the different classes. We considered successively motifs corresponding to the seeds of known micro- RNAs and also all possible motifs of a given length. To rank the motifs, we compared ensemble of decision trees and linear support vector machines. Second, we considered an approach called Segment and Combine that was proposed in [2]. Finally, we also carried out an exhaustive search of all motifs of a given length that satisfy some constraints on specificity and coverage with respect to a given gene category. Results: The ability of the different approaches at identifying relevant motifs was first assessed on genes predicted to be the target of some known miRNAs. In this simple setting, most methods were able to identify the micro-RNA seed. The results obtained on the genes regulated by prolactin 16K are also very encouraging. We were able to identify one micro-RNA already known to play a role in angiogenesis and several motifs are predicted by different approaches as very specific of up- or down-regulation by prolactin 16K. Their relationship with known micro-RNAs is certainly worth exploring. Conclusion: Machine learning approaches are promising techniques for the identification of micro-RNA/gene interactions. Future work will concern the application of the same kind of techniques on promoters for the identification of transcription factor binding sites. [less ▲]

Detailed reference viewed: 73 (16 ULg)
Full Text
Peer Reviewed
See detailPREDetector: A new tool to identify regulatory elements in bacterial genomes
Hiard, Samuel ULg; Marée, Raphaël ULg; Colson, Séverine ULg et al

in Biochemical and Biophysical Research Communications (2007), 357(4), 861-864

In the post-genomic area, the prediction of transcription factor regulons by position weight matrix-based programmes is a powerful approach to decipher biological pathways and to modelize regulatory ... [more ▼]

In the post-genomic area, the prediction of transcription factor regulons by position weight matrix-based programmes is a powerful approach to decipher biological pathways and to modelize regulatory networks in bacteria. The main difficulty once a regulon prediction is available is to estimate its reliability prior to start expensive experimental validations and therefore trying to find a way how to identify true positive hits from an endless list of potential target genes of a regulatory protein. Here we introduce PREDetector (Prokaryotic Regulatory Elements Detector), a tool developed for predicting regulons of DNA-binding proteins in bacterial genomes that, beside the automatic prediction, scoring and positioning of potential binding sites and their respective target genes in annotated bacterial genomes, it also provides an easy way to estimate the thresholds where to find reliable possible new target genes. PREDetector can be downloaded freely at http://www.montefiore.ulg.ac.be/-hiard/PreDetector (c) 2007 Published by Elsevier Inc. [less ▲]

Detailed reference viewed: 87 (31 ULg)
Full Text
See detailPREDetector : Prokaryotic Regulatory Element Detector
Hiard, Samuel ULg; Rigali, Sébastien ULg; Colson, Séverine ULg et al

Poster (2007, February 15)

Background: In the post-genomic area, in silico predictions of regulatory networks are considered as a powerful approach to decipher and understand biological pathways within prokaryotic cells. The ... [more ▼]

Background: In the post-genomic area, in silico predictions of regulatory networks are considered as a powerful approach to decipher and understand biological pathways within prokaryotic cells. The emergence of position weight matrices based programs has facilitated the access to this approach. However, a tool that automatically estimates the reliability of the predictions and would allow users to extend predictions in genomic regions generally regarded with no regulatory functions was still highly demanded. Result: Here, we introduce PREDetector, a tool developed for predicting regulons of DNA-binding proteins in prokaryotic genomes that (i) automatically predicts, scores and positions potential binding sites and their respective target genes, (ii) includes the downstream co-regulated genes, (iii) extends the predictions to coding sequences and terminator regions, (iv) saves private matrices and allows predictions in other genomes, and (v) provides an easy way to estimate the reliability of the predictions. Conclusion: We present, with PREDetector, an accurate prokaryotic regulon prediction tool that maximally answers biologists’ requests. PREDetector can be downloaded freely at http://www.montefiore.ulg.ac.be/~hiard/predetectorfr.html [less ▲]

Detailed reference viewed: 23 (0 ULg)
Full Text
See detailCompiling polymorphic miRNA-target interactions: the Patrocles database.
Hiard, Samuel ULg; Tordoir, Xavier ULg; Coppieters, Wouter ULg et al

Poster (2007, February 15)

Using positional cloning, we have recently identified the mutation responsible for muscular phenotype of the Texel sheep. It is located in the 3’UTR of the GDF8 gene - a known developmental repressor of ... [more ▼]

Using positional cloning, we have recently identified the mutation responsible for muscular phenotype of the Texel sheep. It is located in the 3’UTR of the GDF8 gene - a known developmental repressor of muscle growth - and creates an illegitimate target site for miRNA expressed in the same tissue. This causes miRNA-mediated translation inhibition of mutant GDF8 transcripts which leads to muscle hypertrophy. We followed up on this finding by searching for common polymorphisms and mutations that affect either (i) RNAi silencing machinery components, (ii) miRNA precursors or (iii) target sites. These might likewise alter miRNA-target interaction and could be responsible for substantial differences in gene expression level. They have been compiled in a public database (“Patrocles”: www.patrocles.org), where they are classified in (i) DNA sequence polymorphisms (DSP) affecting the silencing machinery, (ii) DSP affecting miRNA structure or expression and (iii) DSP affecting miRNA target sites. DSP from the last category were organized in four classes: destroying a target site conserved between mammals (DC), destroying a non-conserved target site (DNC), creating a non-conserved target site (CNC), or shifting a target site (S). To aid in the identification of the most relevant DSP (such as those were a target site is created in an antitarget gene), we have quantified the level of coexpression for all miRNA-gene pairs. Analysis of the numbers of Patrocles-DSP as well as their allelic frequency distribution indicates that a substantial proportion of them undergo purifying selection. The signature of selection was most pronounced for the DC class but was significant for the DNC and CNC class as well, suggesting that a significant proportion of non-conserved targets is truly functional. The Patrocles database allowed for the selection of DSP that are likely to affect gene function and possibly disease susceptibility. The effect of these DSP is being studied both in vitro and in vivo. In conclusion, Patrocles-DSP could be widespread and underlie an appreciable amount of phenotypic variation, including common disease susceptibility. [less ▲]

Detailed reference viewed: 29 (2 ULg)
Full Text
Peer Reviewed
See detailPolymorphic miRNA-target interactions : A Novel Source of Phenotypic Variation
Georges, Michel ULg; Clop, Alex; Marcq, Fabienne ULg et al

in Cold Spring Harbor Symposia on Quantitative Biology (2006, June), 71

Studying the muscular hypertrophy of Texel sheep by forward genetics, we have identified an A-to-G transition in the 3'UTRof the GDF8 gene that reveals an illegitimate target site for microRNAs miR-1 and ... [more ▼]

Studying the muscular hypertrophy of Texel sheep by forward genetics, we have identified an A-to-G transition in the 3'UTRof the GDF8 gene that reveals an illegitimate target site for microRNAs miR-1 and miR-206 that are highly expressed in skeletal muscle. This causes the down-regulation of this muscle-specific chalone and hence contributes to the muscular hypertrophyof Texel sheep. We demonstrate that polymorphisms which alter the content of putative miRNA target sites are commonin human and mice, and provide evidence that both conserved and nonconserved target sites are selectively constrained. Wespeculate that these polymorphisms might be important mediators of phenotypic variation including disease. To facilitatestudies along those lines, we have constructed a database (www.patrocles.org) listing putative polymorphic microRNA–targetinteractions. [less ▲]

Detailed reference viewed: 27 (7 ULg)
Full Text
Peer Reviewed
See detailThe callipyge mutation enhances bidirectional long-range DLK1-GTL2 intergenic transcription in cis
Takeda, Haruko ULg; Caiment, Florian ULg; Smit, Maria et al

in Proceedings of the National Academy of Sciences USA (2006), 103(21), 8119-8124

The callipyge mutation (CLPG) is an A to G transition that affects a muscle-specific long-range control element located in the middle of the 90-kb DLK1-GTL2 intergenic (IG) region. It causes ectopic ... [more ▼]

The callipyge mutation (CLPG) is an A to G transition that affects a muscle-specific long-range control element located in the middle of the 90-kb DLK1-GTL2 intergenic (IG) region. It causes ectopic expression of a 327-kb cluster of imprinted genes in skeletal muscle, resulting in the callipyge muscular hypertrophy and its non-Mendelian inheritance pattern known as polar overdominance. We herein demonstrate that the CLPG mutation alters the muscular epigenotype of the DLK1-GTL2 IG region in cis, including hypomethylation, acquisition of novel DNase-I hypersentivite sites, and, most strikingly, strongly enhanced bidirectional, long-range IG transcription. The callipyge phenotype thus emerges as a unique model to study the functional significance of IG transcription, which recently has proven to be a widespread, yet elusive, feature of the mammalian genome. [less ▲]

Detailed reference viewed: 101 (60 ULg)
Full Text
See detailPreDetector : Prokaryotic Regulatory Element Detector
Hiard, Samuel ULg; Rigali, Sébastien ULg; Colson, Séverine ULg et al

Poster (2006, May 17)

PreDetector is a stand-alone software, written in java. Its final aim is to predict regulatory sites for prokaryotic species. It comprises two functionalities. The first one is very similar to Target ... [more ▼]

PreDetector is a stand-alone software, written in java. Its final aim is to predict regulatory sites for prokaryotic species. It comprises two functionalities. The first one is very similar to Target Explorer1. From a set of sequences identified as potential target sites, PreDetector creates a consensus sequence and computes its scoring matrix. This sequence and matrix can be saved on a file and, then, be used to find along a selected genome the sequences that are close enough to the consensus sequence. To this end, a score is attributed to each locus in the genome according to the similarity measure defined by the matrix. The output of this functionality is filtered with a cut-off score and then directly used as input by the second one. The second functionality starts by fetching the gene positions of the selected species from the NCBI server. The loci having above cut-off score are then classified into four classes, allowing multiple classes for one element. This gives the biologists a better view of his discovered sequences. [less ▲]

Detailed reference viewed: 13 (2 ULg)