Meyer Lab

In the listing below, members of my group are indicated in bold. Joint first authors are indicated by an asterisk *.

Pre-prints:

• McCorkindale AL, Wahle P, Werner S, Jungreis I, Menzel P, Shukla CJ, Pereira Abreu RL, Irizarry R, Meyer IM, Kellis M, Rinn JR, Zinzen RP. A gene expression atlas of embryonic neurogenesis in Drosophila reveals complex spatiotemporal regulation of lncRNAs. doi: https://doi.org/10.1101/483461 bioRxiv

  This research is joint work with Robert Zinzen's fruit fly group here at the BIMSB-MDC. Peter Menzel is a post-doc in my group.

• Bogdanow B, Eichelbaum K, Sadewasser A, Wang X, Husic I, Paki K, Hergeselle M, Vetter B, Hou J, Chen W, Wiebusch L, Meyer IM, Wolff T, Selbach M. The dynamic proteome of influenza A virus infection identifies M segment splicing as a host range determinant. doi: https://doi.org/10.1101/438176 bioRxiv

  We here investigate the species-specifity of different influenza A strains. This is joint work the proteomics group of Matthias Selbach and his PhD student Boris Bogdanow at the MDC and the influenza A lab of Thorsten Wolff of the Robert Koch Institute in Berlin. We show that species-specificity is regulated via M segment splicing. My group's Bioinformatics analysis reveals that human-derived and avian-derived influenza A genomes have evolved quite different RNA structure elements overlapping the decisive 3' splice site which turns out to be a key determinant of splicing.

Publications:

• Menzel P, McCorkindale AL, Stefanov SR, Zinzen RP, Meyer IM. Transcriptional dynamics of microRNAs and their targets during Drosophila neurogenesis. RNA Biol. 2018 Dec 24 (published online). doi:10.1080/15476286.2018.1558907. PubMed PMID: 30582411. pubmed

  We here provide a comprehensive in silico analysis of tissue-specific transcriptomes comprising dedicated small and total RNA-seq libraries at two distinct developmental stages during early fly neurogenesis. This enables us to investigate the potential functional roles of individual microRNAs with high spatio-temporal resolution in a genome-wide manner. Our study identifies 74 microRNAs that are significantly differentially expressed between the three cell types and the two developmental stages, predicts target genes of down-regulated microRNAs that show a significant enrichment of their target genes related to neurogenesis and also reveals how microRNAs regulate early fly neurogenesis by targeting transcription factors. Peter Menzel is a post-doc, Stefan Stefanov is a PhD student in my group. This is joint work with the Robert Zinzen's fruit fly group here at the BIMSB-MDC.

• Stefanov SR, Meyer IM Deciphering the Universe of RNA Structures and trans RNA–RNA Interactions of Transcriptomes In Vivo: From Experimental Protocols to Computational Analyses. In: Systems Biology. RNA Technologies, Edited by Rajewsky N, Jurga S, Barciszewski J, Springer, 2018. pdf file via open access by Springer

  The recent years have seen a range of promising, new experimental protocols for investigating RNA structures and trans RNA-RNA interactions of entire transcriptomes in vivo. All of these experimenta strategies, however, require comprehensive computational pipelines for processing and interpreting the large-scale raw data and converting it into evidence for actual RNA structures or trans RNA-RNA interactions. In this invited and peer-reviewed book chapter, my PhD student Stefan Stefanov and I introduce and compare the different strategies and propose ideas for potential future improvements.

• Meyer IM. In silico methods for co-transcriptional RNA secondary structure prediction and for investigating alternative RNA structure expression. Methods. 2017 May 1;120:3-16. PubMed PMID: 28433606. pubmed

  Recent advances on the experimental and computational side have identified a range of intriguing biologically relevant RNA molecules (i.e. transcripts) that exhibit more than a single functional RNA structure throughout their cellular life. This invited and peer-reviewed review paper summarizes computational strategies for successfully identifying these RNA structures and proposes the notion of alternative RNA structure expression to denote that a single transcript can encode several RNA structures which are functionally expressed in distinct, different in vivo settings.

• Mazloomian A, Meyer IM. Genome-wide identification and characterization of tissue-specific RNA editing events in D. melanogaster and their potential role in regulating alternative splicing. RNA Biol. 2015;12(12):1391-401. doi: 10.1080/15476286.2015.1107703. PubMed PMID: 26512413. pubmed

  We analyse tissue-specific high-throughput libraries of D. melanogaster to identify sites RNA editing. For this, we introduce a probabilistic analysis pipeline that utilises large input data and explicitly captures ADAR's requirement for double-stranded regions. Our analysis doubles the number of known RNA editing sites in the fruit fly genome. Our editing sites are 3 times more likely to occur in exons with multiple splicing acceptor/donor sites than in exons with unique splice sites (p-value < 2.10(-15)). Furthermore, we identify 244 edited regions where RNA editing and alternative splicing are likely to influence each other. For 96 out of these 244 regions, we find evolutionary evidence for conserved RNA secondary-structures near splice sites suggesting a potential regulatory mechanism where RNA editing may alter splicing patterns via changes in local RNA structure. Our research identifies a new functional role of RNA editing as mechanism for regulating RNA-structure mediated alternative splicing. This is likely to be of key functional importance in other biological settings such as the human brain. Alborz Mazloomian was a PhD student of mine.

• Lai D, Meyer IM. A comprehensive comparison of general RNA-RNA interaction prediction methods. Nucleic Acids Res. 2016 Apr 20;44(7):e61. doi: 10.1093/nar/gkv1477. PubMed PMID: 26673718. pubmed

  There already exists an abundance of methods that aim to predict specific biological classes of trans RNA-RNA interactions, e.g. miRNA-mRNA interactions. It order to identify truly novel classes of biologically relevant interactions, we require general RNA-RNA interaction prediction methods. This is an invited and peer-reviewed review paper by my former PhD student Daniel Lai and me.

• Zhu JY, Meyer IM. Four RNA families with functional transient structures. RNA Biol. 2015;12(1):5-20. doi: 10.1080/15476286.2015.1008373. PubMed PMID: 25751035. pubmed

  In this comprehensive manuscript, we contribute the first four RNA families comprising more than a single, functional RNA structure to the well-known RNA family data base Rfam. Alice Zhu was an MSc student in my group.

• Lai D, Meyer IM. e-RNA: a collection of web servers for comparative RNA structure prediction and visualisation. Nucleic Acids Res. 2014 Jul;42(Web Server issue):W373-6. doi: 10.1093/nar/gku292. PubMed PMID: 24810851. pubmed

  We here introduce our web-server e-RNA which offers a free and open-access collection of five published RNA sequence analysis tools, each solving specific problems not readily addressed by other available tools. Daniel Lai was a PhD student in my group.

• Lai D*, Proctor JR*, Meyer IM. On the importance of cotranscriptional RNA structure formation. RNA. 2013 Nov;19(11):1461-73. doi: 10.1261/rna.037390.112. PubMed PMID: 24131802. pubmed

  This invited and peer-reviewed review paper summarises the diverse range of experimental and theoretical evidence for co-transcriptional RNA structure formation and proposes a range of ideas how existing, deterministic methods for RNA secondary structure prediction could be potentially improved by taking different aspects of co-transcriptional folding into account. The CoFold paper by the same authors shows that some aspects of co-transcriptional folding can be captured in dedicated models and that this significantly improves the prediction accuracy. Daniel Lai was a PhD student and Jeff Proctor an MSc student in my group.

• Zhu JY*, Steif A*, Proctor JR, Meyer IM. Transient RNA structure features are evolutionarily conserved and can be computationally predicted. Nucleic Acids Res. 2013 Jul;41(12):6273-85. doi: 10.1093/nar/gkt319. PubMed PMID: 23625966. pubmed

  We show that orthologous RNA genes from evolutionarily related organisms not only fold into the same final RNA structure, but that their co-transcriptional folding pathways also share conserved transient RNA structural features. Our study comprises the first comparative analysis of folding pathway prediction programs. Our conclusions are based on 6 data sets with known final and transient RNA structural features, the largest data set of this kind to date. Alice Zhu and Jeff Proctor were MSc students and Adi Steif was an undergraduate research student in my group.

• Proctor JR, Meyer IM. CoFold: an RNA secondary structure prediction method that takes co-transcriptional folding into account. Nucleic Acids Res. 2013 May;41(9):e102. doi: 10.1093/nar/gkt174. PubMed PMID: 23511969. pubmed

  We propose and implement a deterministic RNA secondary structure prediction method, called CoFold, that combines thermodynamic and kinetic considerations. For this, we modify the existing minimum free energy (MFE) method RNAfold and combine it with a model that judges the reachability of potential base-pairing partners during co-transcriptional folding. Our method is the first method of this kind. CoFold effectively depends only on a single free parameter that can be robustly trained. CoFold significantly improves the prediction accuracy of RNAfold, in particular for long sequences over 1000 nt. The method has the same memory and time complexity as RNAfold. Jeff Proctor was an MSc student in my group.

• Steif A, Meyer IM. The hok mRNA family. RNA Biol. 2012 Dec;9(12):1399-404. doi: 10.4161/rna.22746. PubMed PMID: 23324554. pubmed

  The hok/sok toxin-antitoxin system of Escherichia coli plasmid R1 increases plasmid maintenance by killing plasmid-free daughter cells. The hok/sok locus specifies two RNAs: hok messenger RNA, which encodes a toxic transmembrane protein, and sok antisense RNA, which binds a complementary region in the hok mRNA and induces transcript degradation. This post-segregational killing mechanism relies upon the ability of the hok messenger RNA to adopt alternative structural configurations which affect ease of translation and the susceptibility of the molecule to degradation. We have identified several hok mRNA paralogs in the genome of E. coli and Hok protein orthologs in the genomes of Enterobacteria. Using a combination of automated search and extensive manual editing, we have complied the first high-quality multiple sequence alignment for the hok messenger RNA that covers all three experimentally validated hok mRNA structures. Adi Steif was an undergraduate research student in my group.

• Shah SP, Roth A*, Goya R*, Oloumi A*, Ha G*, Zhao Y*, Turashvili G*, Ding J*, Tse K*, Haffari G*, Bashashati A*, Prentice LM, Khattra J, Burleigh A, Yap D, Bernard V, McPherson A, Shumansky K, Crisan A, Giuliany R, Heravi-Moussavi A, Rosner J, Lai D, Birol I, Varhol R, Tam A, Dhalla N, Zeng T, Ma K, Chan SK, Griffith M, Moradian A, Cheng SW, Morin GB, Watson P, Gelmon K, Chia S, Chin SF, Curtis C, Rueda OM, Pharoah PD, Damaraju S, Mackey J, Hoon K, Harkins T, Tadigotla V, Sigaroudinia M, Gascard P, Tlsty T, Costello JF, Meyer IM, Eaves CJ, Wasserman WW, Jones S, Huntsman D, Hirst M, Caldas C, Marra MA, Aparicio S. The clonal and mutational evolution spectrum of primary triple-negative breast cancers. Nature. 2012 Apr 4;486(7403):395-9. doi: 10.1038/nature10933. PubMed PMID: 22495314. pubmed

  We investigate a special tumour type of breast cancers, called triple-negative breast cancers (TNBCs), that is defined by a lack of oestrogen and progesterone receptors and ERBB2 gene amplification. It represents approximately 16% of all breast cancer cases. For this, we investigate 104 individual cases of TNBC. We find that these represent a wide spectrum of genomic evolution ranging from a few coding somatic aberrations in a few pathways to hundreds of coding somatic mutations and propose ways of clustering these into individual tumour clonal genotypes. My former PhD student Rodrigo Goya contributed a major part of the transcriptome analysis (RNA-seq data). This paper is the result of collaboration lead by Sam Aparicio, UBC/BC Cancer Agency, Vancouver, Canada.

• Goya R, Meyer IM, Marra MM. Applications of High-Throughput Sequencing. In: Bioinformatics for High Throughput Sequencing, Springer, 2012. books' web-page by Springer

  This invited and peer-reviewed book chapter summarizes and compares different applications of high-throughput sequencing. Rodrigo Goya was a PhD in my group co-supervised by Marco Marra.

• Lai D, Proctor JR, Zhu JY, Meyer IM. R-Chie: a web server and R package for visualizing RNA secondary structures. Nucleic Acids Res. 2012 Jul;40(12):e95. doi: 10.1093/nar/gks241. PubMed PMID: 22434875. pubmed

  We propose and implement an RNA secondary structure visualisation program called R-chie which we make available via a web-server and a corresponding R-package called R4RNA. R-chie allows to visualise structural information (which may include pseudo-knotted RNA secondary structures as well as mutually exclusive base-pairs) alongside corresponding multiple sequence alignments. The users can also visualise quantitative information on structural and alignment features such as computationally derived base-pairing probabilities or experimentally derived accessibility values. Daniel Lai was a PhD student, Jeff Proctor and Alice Zhu were MSc students in my group.

• Morin RD*, Mendez-Lago M*, Mungall AJ, Goya R, Mungall KL, Corbett RD, Johnson NA, Severson TM, Chiu R, Field M, Jackman S, Krzywinski M, Scott DW, Trinh DL, Tamura-Wells J, Li S, Firme MR, Rogic S, Griffith M, Chan S, Yakovenko O, Meyer IM, Zhao EY, Smailus D, Moksa M, Chittaranjan S, Rimsza L, Brooks-Wilson A, Spinelli JJ, Ben-Neriah S, Meissner B, Woolcock B, Boyle M, McDonald H, Tam A, Zhao Y, Delaney A, Zeng T, Tse K, Butterfield Y, Birol I, Holt R, Schein J, Horsman DE, Moore R, Jones SJ, Connors JM, Hirst M, Gascoyne RD, Marra MA. Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature. 2011 Jul 27;476(7360):298-303. doi: 10.1038/nature10351. PubMed PMID: 21796119. pubmed

  We investigate follicular lymphoma (FL) and diffuse large B-cell lymphoma (DLBCL) which constitute two of the most common non-Hodgkin lymphomas (NHLs). We investigate transcriptome and genome sequencing data from a group of 13 individual DLBCL cases and one FL case in order to identify genes with mutations in B-cell NHLs. Comparisons to data from normal cells allow us to identify 109 genes with multiple somatic mutations. These comprise several genes with roles in histone modification which are analysed in more detail. Our results suggest that disruption of chromatin biology plays a key role in lymphomagenesis. My PhD student Rodrigo Goya contributed to the analysis of transcriptome data which helped to identify key non-synonymous mutations under positive selective pressure. This research is part of a collaboration lead by Marco Marra of the Michael Smith Genome Sciences Centre, Vancouver, Canada.

• Lam TY, Meyer IM. Efficient algorithms for training the parameters of hidden Markov models using stochastic expectation maximization (EM) training and Viterbi training. Algorithms Mol Biol. 2010 Dec 9;5:38. doi: 10.1186/1748-7188-5-38. PubMed PMID: 21143925. pubmed

  We propose and implement two, computationally more efficient algorithms for Viterbi and stochastic EM training. The two new algorithms have the added advantage of being significantly easier to implement than existing algorithms. Both algorithms are also implemented in the HMMConverter software package by the same authors. Tin Yin Lam was an MSc student in my group.

• Wiebe NJ, Meyer IM. Transat - method for detecting the conserved helices of functional RNA structures, including transient, pseudo-knotted and alternative structures. PLoS Comput Biol. 2010 Jun 24;6(6):e1000823. doi: 10.1371/journal.pcbi.1000823. PubMed PMID: 20589081. pubmed

  We propose and implement a new method, called Transat, for detecting evolutionarily conserved helices. This includes helices that are transient, mutually exclusive and that would render the RNA structure pseudo-knotted. Transat is a probabilistic method that employs two probabilistic models of evolution (one to capture how base-pairs evolve over time, one to capture how un-paired nucleotides evolve). It takes as input a multiple sequence alignment and an evolutionary tree linking the sequences of the input alignment and produces as output a ranked list of helices that are assigned log-likelihood scores and p-values. We show in a comprehensive performance evaluation that our method is capable of detecting transient, mutually exclusive and pseudo-knotted features. Our method can thus be used to detect riboswitches which cannot be detected using existing methods for RNA secondary structure prediction as these predict exactly one global RNA secondary structure for a given input sequence/alignment. Nick Wiebe was an MSc student in my group.

• Lam TY, Meyer IM. HMMConverter 1.0: a toolbox for hidden Markov models. Nucleic Acids Res. 2009 Nov;37(21):e139. doi: 10.1093/nar/gkp662. PubMed PMID: 19740770. pubmed

  We introduce a C++-based software package called HMMConverter, that allows users without programming expertise to set up complex hidden Markov models (HMMs) and pair-HMMs. The models and the algorithms that are to be used for generating predictions and for training the model's free parameters are specified via an XML-file. Compared to existing software packages, HMMConverter implements a number of unique features such as (1) the Hirschberg algorithm as memory-efficient alternative to the Viterbi algorithm for pair-HMMs, (2) taking into account prior information on the input sequences and (3) three new algorithms for parameter training (Viterbi, Baum-Welch and stochastic EM training) introduced by us. Tin Yin Lam was an MSc student in my group.

• Schöning JC, Streitner C, Meyer IM, Gao Y, Staiger D. Reciprocal regulation of glycine-rich RNA-binding proteins via an interlocked feedback loop coupling alternative splicing to nonsense-mediated decay in Arabidopsis. Nucleic Acids Res. 2008 Dec;36(22):6977-87. doi: 10.1093/nar/gkn847. PubMed PMID: 18987006. pubmed

  We identify an interlocked feedback loop in Arabidopsis thaliana where two RNA-binding proteins (AtGRP7 and AtGRP8) autoregulate and reciprocally crossregulate their alternative splicing by coupling unproductive splicing to NMD. My group predicts conserved RNA structural features in the intronic regions that are likely to be involved in altering the splicing pattern upon binding of the two RNA-binding proteins. This post-transcriptional feedback loop regulates circadian oscillations in Arabidopsis thaliana. This is joint work with Dorothee Staiger's experimental group at the University of Bielefeld, Germany.

• Pain A, Böhme U, Berry AE, Mungall K, Finn RD, Jackson AP, Mourier T, Mistry J, Pasini EM, Aslett MA, Balasubrammaniam S, Borgwardt K, Brooks K, Carret C, Carver TJ, Cherevach I, Chillingworth T, Clark TG, Galinski MR, Hall N, Harper D, Harris D, Hauser H, Ivens A, Janssen CS, Keane T, Larke N, Lapp S, Marti M, Moule S, Meyer IM, Ormond D, Peters N, Sanders M, Sanders S, Sargeant TJ, Simmonds M, Smith F, Squares R, Thurston S, Tivey AR, Walker D, White B, Zuiderwijk E, Churcher C, Quail MA, Cowman AF, Turner CM, Rajandream MA, Kocken CH, Thomas AW, Newbold CI, Barrell BG, Berriman M. The genome of the simian and human malaria parasite Plasmodium knowlesi. Nature. 2008 Oct 9;455(7214):799-803. doi: 10.1038/nature07306. PubMed PMID: 18843368. pubmed

  This paper presents the results of the international Malaria consortium lead by Matt Berriman at the Wellcome Trust Sanger Institute, Cambridge, UK. My MSc student and I contributed the comparative annotation of the newly sequenced Malaria genome of Plasmodium knowlesi by mapping the known genes of the Malaria genome of Plasmodium falciparum. For this, we implemented a modified version of our comparative gene prediction program Projector that is capable of taking prior information on both genomic sequences into account and whose parameters were trained for these two Malaria genomes. I was a post-doc at the University of Oxford, UK, at that time and Karsten Borgwardt my MSc student.

• Meyer IM. Predicting novel RNA-RNA interactions. Curr Opin Struct Biol. 2008 Jun;18(3):387-93. doi: 10.1016/j.sbi.2008.03.006. PubMed PMID: 18485695. pubmed

  This is an invited and peer-reviewed review paper comparing existing and proposing novel computational strategies for successfully predicting novel types of trans RNA-RNA interactions.

• Yen ZC, Meyer IM, Karalic S, Brown CJ. A cross-species comparison of X-chromosome inactivation in Eutheria. Genomics. 2007 Oct;90(4):453-63. PubMed PMID: 17728098. pubmed

  In Eutheria, X inactivation is initiated by the large noncoding RNA Xist. We contribute a computational, comparative study of evolutionarily conserved RNA structural features for the Xist gene in various eutheria. This analysis is challenging due to the fact that the Xist transcript is long (~17 kb), alternatively spliced and contains tandem repeats some of which are species-specific. We identify two regions that contain rodent-specific, conserved RNA structural features that may play a functional role in Xist regulation. This is joint work with Carolyn Brown's experimental group at UBC, Canada.

• Meyer IM, Miklós I. SimulFold: simultaneously inferring RNA structures including pseudoknots, alignments, and trees using a Bayesian MCMC framework. PLoS Comput Biol. 2007 Aug;3(8):e149. PubMed PMID: 17696604. pubmed

  We introduce and implement a theoretical framework, called SimulFold, that is capable of co-estimating a conserved RNA secondary structure (that may contain pseudo-knots), a multiple sequence alignment and an evolutionary tree. Unlike many existing comparative methods for RNA secondary structure prediction, it does not require a fixed input alignment, thereby resolving a key chicken-and-egg problem. SimulFold employs a non-deterministic Bayesian Markov Chain Monte Carlo rather than an SCFG and is computationally very efficient (new RNA structures can be sampled in linear rather than cubic time). Due to the probabilistic nature of the framework and the co-estimation of key features, SimulFold allows un-precedented and detailed insights into sequence and structure conservation.

• Meyer IM. A practical guide to the art of RNA gene prediction. Brief Bioinform. 2007 Nov;8(6):396-414. PubMed PMID: 17483123. pubmed

  This is an invited and peer-reviewed review paper.

• Meyer IM, Miklós I. Statistical evidence for conserved, local secondary structure in the coding regions of eukaryotic mRNAs and pre-mRNAs. Nucleic Acids Res. 2005 Nov 7;33(19):6338-48. PubMed PMID: 16275783. pubmed

  We use custom computational RNA structure prediction methods in combination with statistical analyses to detect several conserved RNA structural features in pre-mRNAs and mRNAs that are likely to be involved in regulating translation initiation (e.g. mouse caveolin 1 gene) and alternative splicing (e.g. human CFTR gene). This is one of the first studies (1) to identify conserved, local RNA structural elements overlapping splice sites, (2) to provide a statistical link between synonymous exon mutations and changes of the splicing efficiency and (3) to show that these changes are likely to be due to changes of the RNA structure that are induced by synonymous mutations.

• Miklós I, Meyer IM. A linear memory algorithm for Baum-Welch training. BMC Bioinformatics. 2005 Sep 19;6:231. PubMed PMID: 16171529. pubmed

  We introduce a new mathematical algorithm for Baum-Welch parameter training that is computationally more efficient than existing ones and also significantly easier to implement. This result is relevant to all applications that employ hidden Markov models (HMMs) (and their variants) in conjunction with Baum-Welch parameter training.

• Miklós I, Meyer IM, Nagy B. Moments of the Boltzmann distribution for RNA secondary structures. Bull Math Biol. 2005 Sep;67(5):1031-47. PubMed PMID: 15998494. pubmed

  We propose and implement the first algorithm for calculating arbitrary moments of the Boltzmann distribution of RNA secondary structures. Using our new algorithm, we find that biological RNAs have a Boltzmann distribution that comprises an ensemble of structures that are close to the minimum free energy structure. This feature is likely to convey an evolutionarily advantages to biological RNAs and is absent from randomly generated RNAs with the same overall sequence properties.

• Pedersen JS*, Meyer IM*, Forsberg R, Simmonds P, Hein J. A comparative method for finding and folding RNA secondary structures within protein-coding regions. Nucleic Acids Res. 2004 Sep 24;32(16):4925-36. PubMed PMID: 15448187. pubmed

  We introduce a comparative method, called RNA-Decoder, that is capable of detecting conserved RNA secondary structures in RNAs that may be partly protein-coding (e.g. viral RNA+ genomes, pre-mRNAs, mRNAs). The method employs new evolutionary models [5] to capture different, overlapping evolutionary constraints and, unlike Pfold, explicitly captures also local rather than only global RNA secondary structure. We show that RNA-Decoder outperforms existing methods for RNA secondary structure prediction that do not explicitly capture the protein-context (e.g. RNAalifold, Pfold, Mfold). RNA-Decoder was, for example, used for the genome-wide structural annotation of the HIV genome (paper by Kevin Week's group, Nature (2009)). RNA-Decoder is still unique in the sense that it explicitly captures the known protein-coding context of RNAs.

• Meyer IM, Miklós I. Co-transcriptional folding is encoded within RNA genes. BMC Mol Biol. 2004 Aug 6;5:10. PubMed PMID: 15298702. pubmed

  We show that structural RNA genes not only encode information on their known, final RNA secondary structure, but also information on their co-transcriptional folding pathway. More specifically, we show that transient RNA structures that are likely to prevent the co-transcriptional formation of the final RNA structure are suppressed, whereas transient RNA structures that could help the formation of the final RNA structure are encouraged. This paper was featured as a special highlight of BMC Molecular Biology.

• Pedersen JS*, Forsberg R*, Meyer IM, Hein J. An evolutionary model protein-coding regions with conserved RNA structure. Mol Biol Evol. 2004 Oct;21(10):1913-22. PubMed PMID: 15229291. pubmed

  We propose and implement several probabilistic models of evolutionary that model conserved RNA structural features overlapping known protein-coding regions. The key difficulties we address are (1) to capture the two different evolutionary constraints, (2) to propose a way of avoiding long-range correlations due to the coding-context and (3) to parametrise the evolutionary models in ways that capture the key sequence and structure signals while also allowing for robust parameter training given the scarcity of our training data.

• Meyer IM, Durbin R. Gene structure conservation aids similarity based gene prediction. Nucleic Acids Res. 2004 Feb 4;32(2):776-83. PubMed PMID: 14764925. pubmed

  We propose and implement a comparative gene prediction method, called Projector, that maps known genes of one genome to orthologous regions of a related genome. Similar to Doublescan, Projector employs a pair hidden Markov model (pair-HMM) and aligns and predicts pairs of genes simultaneously. We also incorporate a heuristic algorithm that allows us to generate predictions in near-linear time. We show that Projector outperforms protein-based methods such as Genewise, especially for pairs of genes that are more distantly related.

• Meyer IM, Durbin R. Comparative ab initio prediction of gene structures using pair HMMs. Bioinformatics. 2002 Oct;18(10):1309-18. PubMed PMID: 12376375. pubmed

  We propose a new comparative method for ab-initio gene prediction, called Doublescan, that aligns and predicts pairs of orthologous genes from mouse and human simultaneously, thereby avoiding the need for a fixed high-quality input alignment that other comparative methods require. The method is special in that it can also handle pairs of orthologous genes that are related by exon-fusion and exon-splitting. These account for 16% of orthologous mouse-human gene pairs.

• Bentvelsen S, Meyer I. The Cambridge jet algorithm: features and applications, European Physics Journal C4 (1998) 74.

  We here introduce a new mathematical algorithm for jet detection in high-energy particle physics. This is research done by me while being an MSc student at CERN in Geneva, Switzerland.