UCSC Genome Browser

Complete human genome

Transposable Elements
45% of the human genome is occupied by transposons and transposon-like

repetitive elements.
Barbara McClintock (1902-1992) in 50s.
Nobel prize in 1983

Схожесть с ретровирусами
Retrovirus reverse transcripiton
http://www.youtube.com/watch?v=eS1GODinO8w
Class II
Class I – retrotransposons (via RNA

intermediate)

Active
Non-Active

First Layer of Genome Annotation

Epigenetics

Second Layer of Genome Annotation

Third Layer of Genome Annotation

Data Accumulation

ENCODE: Encyclopedia of DNA Elements

Digital Universe
Like the Physical Universe the Digital Universe is also expanding but

much faster doubling every two years – and by 2020 will be 44 zettabytes (10^ 21)
Every second a new 205 000 bytes come to being
At the end of this lecture the digital universe will grow by 2 214 000 000 bytes or 2.2 GB.

Слайд 25

Digital Universe
Data Universe Will Expand To 44 Trillion GBs By 2020

Слайд 26

Что делать?

Слайд 27

Что получилось? (Success Stories)

Слайд 28

СКРЫТЫЕ ЦЕПИ МАРКОВА

Слайд 29

Gene Prediction

Слайд 30

E0
E1
E2
E2
E1
E0
N
P
Eterm
P
Einit
polyA
5’ UTR
I0
I1
I2
I0
I1
I2
Esngl
Esngl
Einit
Eterm
forward strand
backward strand
3’ UTR
5’ UTR
3’ UTR
polyA
E- exons
I- introns
single exon
5’ UTRs

3’ UTRs
P- promoter region polyA site N- intergenic region

GeneMark HMM

Слайд 31

Promoter prediction
Hidden Markov model with six interpolated Markov chain submodels
upstream 1

and 2,
TATA box, spacer,
Initiator
downstream.
Gaussian densities of DNA physicochemical properties.
Neural network classifier

McPromoter

Слайд 32

predict tissue-dependent changes in alternative splicing for thousands of exons.
1,014 features:

known motifs, new motifs, short motifs and features describing transcript structure
trained on RNA-seq data
single-layer logistic Bayesian network or neural network, or a weighted combination of single-layer decision trees.

Nature 2010

Слайд 33

Genome intrinsic organization can explain ,50% of the in vivo nucleosome positions
Probabilistic

nucleosome–DNA interaction model - built on dinucleotide distrubution
Thermodynamic model for predicting nucleosome positions genome-wide.

Слайд 34

Слайд 35

Schematic overview of epigenetic regulatory mechanisms.
Yonggang Zhou et al. Circ Res.

2011;109:1067-1081

Слайд 36

Random Forest model predicts cancer mutation densities from epigenomic mark ups

Слайд 37

Слайд 38

We have many experimental genome-wide annotations
Question 1: Are different annottaions correlated? To

what extent?

Question 2: Can we find patterns in annotations?
(Unsupervised learning)

Слайд 39

Annotations under different conditions

Слайд 40

Как много данных?
Roadmap Epigenomics
~ 3 000 полногеномных данных
ENCODE Encyclopedia of Genomic

Elements
~ 9000 полногеномных данных
International Cancer Genome Consortium
~ 20 000 patients (~50 типов рака)
The Cancer Genome Atlas
~ patients 11 000 (~33 типа рака)

Слайд 41

Открытые вопросы
Какие участки кода работают одновременно?
Как переключать режимы работы клетки?
Как перепрограммируется

код для разных типов тканей?
Сколько механизмов регуляции существует в клетках (надежда на универсальность)?

UCSC Genome Browser

Содержание