Lecture 7 gene prediction

Question 1

Once a genome seq info has been successfully sequenced and assembled, what type of approach is used to predict its gene structure?

Answer 1

computational approaches

Question 2

What are computational gene predictions and what do they include?

Answer 2

Computational gene prediction is necessary to obtain comprehensive functional information on genes and genomes. The process includes detection of the location of open reading frames (ORFs).

Question 3

What else does the computational gene prediction require in eukaryotes?

Answer 3

Description of the structures of introns/exons.

Question 4

What is the main goal of gene prediction?

Answer 4

The main goal is to describe all the genes computationally with 100% accuracy.

Question 5

What aren’t conserved in coding regions and how does it effect gene prediction?

Answer 5

Motifs aren’t conserved in coding regions making gene prediction one of the most difficult problems in the field of pattern recognition.

Question 6

What are the 3 way for finding genes in genomes?

Answer 6

1) Similarity-based or Comparative

2) Ab initio = “from the beginning”

3) Combined “evidence-based” (BEST)

Question 7

1) Similarity-based or Comparative

Answer 7

• BLAST - Do other organisms have similar sequence?
  (Is sequence similar to known gene or protein)

Question 8

2) Ab initio

Answer 8

• Ab initio meaning, “from the beginning” predicts without explicit comparison with cDNA or proteins via
  “rule-based” gene models - but rules are derived from statistical analysis of datasets

Question 9

3) Combined “evidence-based”

Answer 9

• Combine gene models with alignment to known ESTs & protein sequences

Question 10

What is the gene density in prokaryotes

Answer 10

High, more than 90% of their genome contains coding seq w very few repetitive sequences

Question 11

What is the gene prediction in prokaryotes?

Answer 11

Each prokaryotic gene is composed of a single contiguous stretch of ORF coding for a single protein or RNA with no interruptions within a gene.
- bac genes have a start codon ATG. GTG and TTG are used as alternative start codons at times.
- At the end of the protein coding region is a stop codon, TAA, TAG, TGA

Question 12

As there may be multiple ATG, GTG, or TGT codons in a frame in prokaryotes, how can the start codon be located?

Answer 12

-Identification of the ribosome binding site (Shine-Delgarno sequence) can help locate the start codon. The ribosomal binding site is located slightly upstream of the translation start codon and has a consensus motif of AGGAGGT.
- Identification o the stop Condon is straight forward

Question 13

What is the ribosomal binding site/ Shine-Delgarno sequence?

Answer 13

a stretch of purine-rich sequence complementary to 16S rRNA in the ribosome.

Question 14

How can potential coding regions be detected?

Answer 14

by looking for ORFs

Question 15

What kind of ORF should be used and how can a purposed frame be confirmed for the presence of the gene?

Answer 15

• Long open reading frames may be a gene
• A basic approach is to scan for ORFs whose length exceeds certain threshold (60 amino acids/180 nucleotides)
  – A proposed frame can be confirmed by the presence of other signals such as the Shine–Delgarno sequence.

Question 16

When should a stop codon be seen at random?

Answer 16

one stop codon every (64/3) = 21 codons

Question 17

what is a disadvantage of using stop codons in an ORF to detect a gene

Answer 17

genes are usually longer than 21 codons therefore if stop codons are used a whole gene may not be identified.

Question 18

What is the disadvantage of using using certain thresholds to scan ORFs

Answer 18

some genes (e.g. some neural and immune system genes) are relatively short, therefore using a long ORF would be inaccurate.

Question 19

What is th strongest indicator of a protein-coding frame.

Answer 19

Detection of homologs from searching long ORFs against database of confirmed genes using BLAST

Question 20

what can testing the GC bias in ORF exam?

Answer 20

the non-randomness of nucleotide distribution in ORFs.

Question 21

in a coding seq what nucleotides is usually in third position?

Answer 21

G or C

Question 22

How can GC bias be used to identify a ORF

Answer 22

By plotting the probability of GC composition in the 3rd position, regions with values significantly above the random level can be identified, we are searching for ORFs that have higher levels of G/C at the third codon position relative to what we would see by chance alone. We are looking for enrichment.

Question 23

how can Condon usage be used to test an ORF

Answer 23

By creating a 64-element hash table and counting the frequencies of codons in an ORF

Question 24

How does the uneven usage of codons in nature compensate for pitfalls of the ORF length test?

Answer 24

Amino acids typically have more than one codon, but in nature certain codons are preferred, therefore we test for ORFs that have the “likely” codon usage, this compensates for pitfalls of the ORF length test as an ORF is more “believable” than another if it has more “likely” codons.

Question 25

Markov Models

Answer 25

A Markov model explains the likelihood of the arrangement of nucleotides in a DNA sequence, where the probability of a specific position in the sequence is influenced by the preceding k positions.

Question 26

Markov models are a well known tool for analysing sequence data and are used by ?

Answer 26

GeneMark and Glimmer.

Question 27

By looking at proceeding bases what does fixed order Markov models predict?

Answer 27

each base of a sequence

Question 28

Zero- order Markov model

Answer 28

Assumes each base occurs independently with a given probability.

Question 29

1st-order Markov model

Answer 29

looks at the preceding base to determine what base will follow.

Question 30

2nd- order Markov model

Answer 30

looks at the preceding two bases to determine what base will follow.

Question 31

What fact do Markov models exploit

Answer 31

Markov models exploit the fact that oligonucleotide distributions in coding regions are different from those for the noncoding regions.

Question 32

what does a longer oligomer unit indicate?

Answer 32

The longer the oligomer unit, the more non-randomness can be described for the coding region.

Question 33

The higher the order of a Markov model….

Answer 33

the more accurately it can predict a gene.

Question 34

How are Markov Models built

Answer 34

Markov models are built in sets of three nucleotides, describing non- random distributions of trimers or hexamers, etc.

Question 35

What are the parameters of Markov models trained with?

Answer 35

set of sequences with known gene locations.

Question 36

What can the accuracy of a prediction programme be evaluated with?

Answer 36

parameters such as sensitivity and specificity.

Question 37

Sensitivity

Answer 37

the ability to include correct predictions.

Question 38

Specificity

Answer 38

ability to exclude incorrect predictions.

Question 39

When is a program considered accurate?

Answer 39

both sensitivity and specificity are simultaneously high and approach a value of 1.

Question 40

whe sensitivity is high but specificity is low?

Answer 40

the program is said to have a tendency to overpredict.

Question 41

When sensitivity is low but specificity high,

Answer 41

the program is too conservative and lacks predictive power.

Question 42

What four features are used to describe the concept of sensitivity and specificity accurately?

Answer 42

• True Positive (TP), which is a correctly predicted feature/gene.
• False Positive (FP), which is an incorrectly predicted feature/gene.
• False Negative (FN), which is a missed feature/gene.
• True Negative (TN), which is the correctly predicted absence of a feature/gene.

Question 43

Sensitivity is given by?

Answer 43

Sn = TP/(TP + FN).

Question 44

Specificity is given by

Answer 44

Sp = TP/(TP + FP).