Розробка та застосування методів in silico для визначення сайтів зв’язування транскрипційних факторів у еукаріот на прикладі ISGF-3

Ключові слова: інтерферон альфа, сайт зв'язування транскрипційного фактора, ISRE, ISGF-3, позиційно-вагова матриця, приховані моделі Маркова.

Summary

Tokovenko B. T. Design and implementation of the in silico methods to identify eukaryotic transcription factor binding sites: ISGF-3 case study. - Manuscript.

Thesis for a Philosophy Doctor (PhD) degree in Biology, speciality 03.00.03 - molecular biology. - Institute of Molecular Biology and Genetics of National Academy of Sciences of Ukraine, Kyiv, 2010.

Thesis focuses on identifying genes of primary response to interferon alpha (IFNб) with bioinformatics methods, as well as developing improved methods and tools for the computational identification of transcription factor binding sites (TFBS).

Initial application of the TFBS search method based on position weight matrices (PWM) for the identification of the interferon-stimulated response element (ISRE), which is specifically bound by interferon-stimulated genes factor 3 (ISGF-3), in the Rattus norvegicus genome proved that it is possible to obtain meaningful results (or, at the very least, results matching expectations built on prior knowledge). However, high percentage of false-positive results significantly hinders results analysis: with a 0.8 similarity score threshold of the M00258 matrix from the TRANSFAC 7.0 Public database a total of 5214 ISRE-like sequences were found in 4571 promoters; 768 of the genes (16.8%) made it through the next step of a custom phylogenetic footprinting filtering (i.e. they had ISREs with conservative positions relative to the transcription start sites of the orthologous genes in both R. norvegicus and Mus musculus). Analysis of the list of genes using Gene Ontology (GO) indicated enrichment of the categories belonging to the known IFNб target systems: humoral immune response, response to virus, positive regulation of B cells differentiation, and others. Unexpectedly, the nervous system development category, which is not a known target of IFNб, was also found on the list of enriched categories.

In order to improve specificity of the TFBS search a new method based on hidden Markov models (HMMs) was developed. While PWM-based search methods assume that neighboring nucleotide frequencies are fully independent of each other, the new HMM-based method allows preserving these dependencies. After training the HMM-based model on the known nucleotide sequences of the specific TFBS, new method allows searching for similar sites with no assumptions on the independence of each binding site position. In order to simplify the use of the new method, an automatic similarity score threshold estimation procedure was developed, which attempts to estimate optimal threshold value for each set of provided known TFBS sequences.

Comparison of the methods based on PWM and HMM while searching for ISRE in the 2^nd and 3^rd exons of the protein-coding genes of R. norvegicus (where one would expect close to zero functional ISREs) revealed a significant difference in the frequency distribution of the similarity scores. While for the PWM-based method similarity scores were normally distributed with arithmetic mean 0.3, the HMM-based method produced a logarithmic distribution with л = 0.16. Thus, in accordance with initial expectations, the HMM-based TFBS search method has higher search specificity, than PWM-based methods.

Using the new HMM-based method, own implementation of the classical PWM-based method, and a custom implementation of phylogenetic footprinting method (based on the comparison of the orthologous genes' promoters), a free-to-use online tool called COTRASIF (conservation-aided transcription factor binding site finder) was developed. The new tool features ease of use and 20 eukaryotic promoter collections, which are regularly updated from the primary Ensembl database of genome annotations with the help of the custom-built automatic promoters import pipeline. Gene orthology information is also regularly updated from Ensembl database. Automatic updates ensure that COTRASIF will stay up-to-date for a longer period of time. TFBS matrices from both JASPAR and TRANSFAC databases are also integrated for the ease of use. Based on requests from COTRASIF users, a standalone version for Linux (32/64-bit) was made available on the COTRASIF website.

For IFNб, the dominating Jak-STAT signal transduction pathway is well-studied. It is known that ISGF-3 is the final activated factor, which specifically binds with ISREs in the promoters of target genes. Thus, identification of the ISGF-3 binding sites should help identify the genes of primary response to IFNб.

We performed a search for ISREs with the HMM-based method in the promoters of R. norvegicus and M. musculus genes obtained from Ensembl version 52. We found 743 putative ISREs in 707 R. norvegicus genes of putative primary response to IFNб. Phylogenetic footprinting reduced the list to 162 genes with evolutionary-conserved ISREs.

Gene list GO analysis with FatiGO identified a single significantly enriched (p=1,23•10^-4, corrected p=8,52•10^-3) category immune response (GO:0006955) with 9 genes in it. Out of 9 genes, 5 are known IFNб targets, 2 have M. musculus orthologs with experimental evidence of their regulation by IFNб, and for 2 genes (Cxcl4, Mbl1) no data on the correlation of their expression with IFNб stimulation was found.

Thorough comparison of the 162 genes with literature data revealed that for 61 genes there are already (both direct and indirect) experimental indications of their regulation by IFNб. For the remaining 101 genes no evidence of IFNб regulation was found. GO analysis of the 101 genes identified a single enriched category synapse component, containing 3 genes: Pick1 (PRKCA-binding protein), Grin3a (NMDA receptor subunit 3A), and Gabra2 (GABA_A receptor subunit alpha-2). Confirmation of their regulation by IFNб will indicate the discovery of a new function of IFNб, namely regulation of the stimulating synapses functions, axon growth, and synaptogenesis. It is well-known that IFNб treatment of some viral infections and cancer types results in side-effects, including headaches, depression etc. Our results for the first time point at likely nervous system gene targets of IFNб, and enable further molecular studies of the new IFNб effects.

In conclusion, the new HMM-based search method, the COTRASIF tool, TFBS search results prospectivity criteria, and GO analysis revealed 24 novel putative targets of IFNб regulation.

Keywords: interferon alpha, transcription factor binding site, ISRE, ISGF-3, position-weight matrix, hidden Markov models.

Аннотация

Токовенко Б.Т. Разработка и применение методов in silico для определения сайтов связывания факторов транскрипции у эукариот на примере ISGF-3. - Рукопись.

Диссертация на соискание ученой степени кандидата биологических наук по специальности 03.00.03 - молекулярная биология. - Институт молекулярной биологии и генетики НАН Украины, Киев, 2010.

Разработан улучшенный метод поиска сайтов связывания факторов транскрипции (ССФТ) на основе скрытых моделей Маркова. Показано, что разработанный метод обеспечивает более высокую специфичность поиска ССФТ по сравнению с методом на основе позиционно-весовых матриц. Разработан общедоступный, бесплатный, простой в использовании онлайн-инструмент COTRASIF для нахождения в промоторах генов эукариот ССФТ в масштабах геномов. С помощью инструмента COTRASIF впервые обнаружено 162 гена первичного ответа на действие интерферона альфа (ИФНб). Установлено, что для 61 гена первичного ответа на ИФНб уже есть экспериментальные данные, которые указывают на их регуляцию ИФНб. Впервые показано, что ряд генов (Pick1, Grin3a, Gabra2), которые отвечают за развитие и функционирование нервной системы, также могут регулироваться ИФНб. Разработана система оценки перспективности результатов поиска ССФТ, которая использует 3 критерия. Определены 24 гена-кандидата первичного ответа на ИФНб для их первоочередной экспериментальной проверки.

Ключевые слова: интерферон альфа, сайт связывания фактора транскрипции, ISRE, ISGF-3, позиционно-весовая матрица, скрытые модели Маркова.

Размещено на Allbest.ru