Napredna pretraga

Pregled bibliografske jedinice broj: 500679

Detecting natural selection for translational efficiency in bacterial and archaeal genomes: a machine learning approach


Supek, Fran
Detecting natural selection for translational efficiency in bacterial and archaeal genomes: a machine learning approach 2010., doktorska disertacija, Prirodoslovno-matematički fakultet, Zagreb


Naslov
Detecting natural selection for translational efficiency in bacterial and archaeal genomes: a machine learning approach

Autori
Supek, Fran

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, doktorska disertacija

Fakultet
Prirodoslovno-matematički fakultet

Mjesto
Zagreb

Datum
18.10

Godina
2010

Stranica
149

Mentor
Tomislav Šmuc

Ključne riječi
Codon usage bias; Random Forest; Support Vector Machines; non-coding DNA; highly expressed genes; translational selection; adaptome

Sažetak
Unequal use of synonymous codons in prokaryotes is largely a consequence of mutational biases, but highly expressed genes exhibit a preference towards codons that enable more efficient translation. I have tested and implemented in a freely accessible software package (INCA) several distance-based approaches to codon usage analysis and proposed a novel method (MILC) robust to changes in gene length and composition. Additionally, I have introduced a classifier-based computational framework that can distinguish between the two principal influences on codon usage. Evidence for translational selection was found to be universal in prokaryotic genomes, and the extent of such selection was quantified. I have performed extensive statistical testing for enrichment of specific gene functional categories with codon-optimized genes. Since presence of codon optimizations can be used as a proxy for expression levels in every sequenced genome, I have contrasted predicted gene activity within groups of Bacteria and Archaea defined by phenotype, environment or genotype, outlining microbial 'adaptomes'. Additionally, I have examined the interplay of codon usage and 5' mRNA secondary structures in determining the success of heterologous gene expression in an E. coli host.

Izvorni jezik
Engleski

Znanstvena područja
Biologija



POVEZANOST RADA


Projekt / tema
098-0000000-3168 - Strojno učenje prediktivnih modela u računalnoj biologiji (Tomislav Šmuc, )

Ustanove
Institut "Ruđer Bošković", Zagreb

Autor s matičnim brojem:
Fran Supek, (268331)