Transcript 6. ORF Calling

ORF Calling
ORF Calling

Why?

Need to know protein sequence

Protein sequence is usually what does the work

Functional studies


Crystallography

Proteomics
Similarity studies


Proteins are better for remote similarities than
DNA sequences
Protein sequences change slower than DNA
sequences
ORF Calling
Extrinsic gene calling
Compare your DNA sequences to known
sequences. Needs other sequences that
are known!
Intrinsic gene calling
Only use information in your DNA
sequences. Does not use other
information.
Extrinsic gene calling

Start with DNA sequence

Translate in all 6 reading frames
Why are there 6 reading frames?
3
2
1
AG TAA AAC TTT AAT TGT TGG TTA A
A GTA AAA CTT TAA TTG TTG GTT AA
AGT AAA ACT TTA ATT GTT GGT TAA
AGT AAA ACT TTA ATT GTT GGT TAA
||| ||| ||| ||| ||| ||| ||| |||
TCA TTT TGA AAT TAA CAA CCA ATT
-1
-2
-3
TCA TTT TGA AAT TAA CAA CCA ATT
TC ATT TTG AAA TTA ACA ACC AAT T
T CAT TTT GAA ATT AAC AAC CAA TT
Extrinsic gene calling

Start with DNA sequence

Translate in all 6 reading frames


Compare your sequence to known protein
sequences
Find the ends of each, and call those genes!
For example
Protein encoding
gene
DNA
sequence
}
Similar
protein
sequences
e.g. from BLAST
Uses of extrinsic calling


This is how (most) metagenome ORF calling is
done
Eukaryotic ORF calling – especially using EST
sequences
Problems with extrinsic calling

Very slow (depending on search algorithm)

Dependent on your database

Only finds known genes
Alternatives to extrinsic gene calling

Intrinsic gene calling

Ab initio gene calling

What are the start codons?

What are the stop codons?
ATG
TAA TAG TGA
How frequently do stop codons appear?
Approximately once every 20 amino acids at
random!
A stretch of 100 amino acids is likely to have a
stop codon!
How to call ORFs (the easy way)
3
2
1
DNA
-1
-2
-3
Find all the stop codons
3
2
1
DNA
-1
-2
-3
Find all the ORFs > x amino acids
X is often 100 amino acids
3
2
1
DNA
-1
-2
-3
Trim to those ORFs that have a start
3
2
1
DNA
-1
-2
-3
Remove “shadow” ORFs
Short ORFs that overlap others
3
2
1
DNA
-1
-2
-3
Trim the start sites to first ATG
3
2
1
DNA
-1
-2
-3
These are the ORFs
3
2
1
DNA
-1
-2
-3
Intrinsic ORF calling using
Markov Models
Markov Models


Based on language processing
Common for gene and protein finding,
alignments, and so on
What is the most common word?
English: the
Spanish: el (la)
Portuguese: que
Scrabble
Scrabble
In scrabble, how do they score the letters?
The most abundant letters (easiest to place on
the board) are given the lowest score
Scrabble
1 point: E, A, I, O, N, R, T, L, S, U
2 points: D, G
3 points: B, C, M, P
4 points: F, H, V, W, Y
5 points: K
8 points: J, X
10 points: Q, Z
Frequency of letters
Making up sentences
If I want to make up a sentence, I could choose
some letters at random, based on their
occurrence in the alphabet (i.e their scrabble
score)
rla bsht es stsfa ohhofsd
Lets get clever!
What follows a period (“.”)?
Usually a space “ ”
What follows a t?
Usually an “i” (-tion, -tize, ...)
Frequency of two letters
When the first letter is “t” (from 3,269 words):
ti
51%
te
20%
ta
15%
th
8%
Level 1 analysis
Choose a letter based on the probability that it
follows the letter before:
sha nd
t uc t hi ney
me
l e ol l d
Levels of analysis
1 letter (a, e, o …)
Zero order model
2 letters (th, ti, sh …)
First order model
3 letters (the, and, …)
Second order model
4 letters (that, …)
Third order model
Markov models
With about 10th order Markov models of English
you get complete words and sentences!
Markov models
With about 10th order Markov models of English
you get complete words and sentences!
Scoring words with Markov Models

If I choose random letters how can I tell if they
are real words?
Sum the scores of 10th order Markov models
across the words … if it is high it is likely to be a
real word!
In reality, maybe use 1st, 2nd, 3rd, 4th, 5th, 6th …
order models and compare to some known
words
Markov Models and ORF calling
Codons have three letters (ATG, CAC, GGG, ...)
Use a 2nd order Markov model for ORF calling
The frequency of a letter is predicted based on
the frequency of the two letters before
Scrabble
Scrabble (México)
Do English and Spanish use the same letters?
Scrabble (México)
Scrabble (US)
1 point: E, A, I, O, N, R, T, L, S, U
2 points: D, G
3 points: B, C, M, P
4 points: F, H, V, W, Y
5 points: K
8 points: J, X
10 points: Q, Z
Based on the front page of the NY Times!
Scrabble (Spanish)
1 point: A, E, O, I, S, N, L, R, U, T
2 points: D, G
3 points: C, B, M, P
4 points: H, F, V, Y
5 points: CH, Q
8 points: J, LL, Ñ, RR, X
10 points: Z
What about scrabble scores for DNA?
Will vary with the composition of the organism!
Remember, some organisms have high G+C
compared to A+T
Markov Models and ORF calling
Use a 2nd order Markov model for ORF calling
The frequency of a letter is predicted based on
the frequency of the two letters before
Problems!
Need to train the Markov model – not all
organisms are the same
Can use phylogentically close organisms
Can use “long orfs” – likely to be correct because
unlikely to be random stretches without a stop
codon!
Interpolated Markov Model
(The imm in GLIMMER)
Markov Models order 1-8 (word size2-9
2-9)
Discard (or ↓ weight) for rare words
Promote (or ↑ weight) for common words
Probability is the sum of all probabilities from 1-8
RNA genes
As with proteins, two main methods:
Ab initio
• Intrinsic
Homology based
• extrinsic
Ribosomes
Ribosomes are made of proteins and RNA
30S subunit from Thermus
aquaticus
Blue: protein
Orange: rRNA
E. coli
16S rRNA
secondary
structure
Variable region
Conserved
region
V5
(28,
29)
V6
(37
)
V7
(43)
V8
(45,
46)
V4
(P231,
24)
Variable regions in
the 16S rRNA.
Vn – 9 regions
(n) – variable loop(s)
forward/rev primers
V9
(49)
V3
(18)
Van de Peer Y, Chapelle S, De Wachter R.
(1996) A quantitative map of nucleotide
substitution rates in bacterial rRNA.
Nucl. Acids Res. 24:3381-3391
V1
(6)
V2 (811)
Ribosomes
Ribosomes are made of proteins and RNA
Prokaryotic ribosome:
Large subunit:
50S
5S and 23S rRNA genes
Small subunit:
30S
16S rRNA gene
Finding 16S genes
Easiest way is iterative:

BLAST

ALIGN

TRIM
Problem: secondary structure makes
identification of the ends difficult
Finding tRNA genes
Not as easy as rRNA
Much shorter
Varied sequence
Only conservation is 2° structure
tRNAScan-SE
Sean Eddy
Use it!
tRNA-Phe by Yikrazuul - Own work.
Licensed under CC BY-SA 3.0 via Wikimedia Commons
https://commons.wikimedia.org/wiki/File:TRNA-Phe_yeast_en.svg
How does this relate to tRNA?
tRNA structure

Start of acceptor stem (7-9 bp)

D-loop (4-6-bp) stem plus loop

anticodon arm (6-bp) stem plus loop with
anticodon

T-loop (4-5-bp) stem plus loop

End of acceptor stem (7-9 bp)

CCA to attach amino acid (may not be in
sequence ... added during processing)