analysis of gene expression 2

Question 1

identification of differentially expressed genes

Answer 1

• data analysis methods:
  
  • fold change/thresholds
  • t-test

Question 2

fold change

Answer 2

• average fold changes of experimental replicated
  
  • or average of log ratios (more accurate)
• decide threshold
• modification = additional criteria for intensity change
  
  • require absolute change e.g. by 10 units
  • or floor intensity data (all below 10 to 0)
  • reduces number of false positives and corrects for systematic errors
  • fewer hypotheses to be tested

Question 3

fold change advantages

Answer 3

• simple to implement and easy to calculate
• straightforward interpretation
• can use with few replicates

Question 4

fold change disadvantages

Answer 4

• small intensity changes can produce large calculated fold changes in poorly expressed genes
• doesn’t account for noisy data
  
  • outliers have large effect on average fold change
• not statistically-based
  
  • threshold?
  • convenient not mathematical

Question 5

t test analysis

Answer 5

• statistical version of fold change
• assumes 2 samples are normally distributed
• investigates whether their means are the same or different by calculating t statistic for each replicate
• null hypothesis - average expression is same for both samples

Question 6

multiple correction testing

Answer 6

• more tests means you increase the number of apparently significant results you would expect by chance alone
• if alpha = 0.05, you would expect 50 unusual results in 1000 tests
• need to correct for this

Question 7

t test

advantages and disadvantages

Answer 7

• advantages:
  
  • statistical
  • fewer false positives than fold change
  • can combine RNAseq and microarray data
• disadvantages:
  
  • usually few replicates - limits statistical power
  • can lead to large gene-to-gene fluctuations in calculated standard deviation with small replicate number

Question 8

DNA binding sites

Answer 8

• indicate how translaiton is controlled
• can predict regions that lead to expression of particular genes
• experimental identification or de novo prediction to create binding site library

Question 9

using DNA motif knowledge

Answer 9

• search sequence for known sites
• identify and search for restriction sites
• use information to model binding site
• create consensus
  
  • decide number of allowed mismatches
  • depends on sequence properties
• create weight matrix
• create position frequency matrix
• create position weight matrix

Question 10

binding site PWM

Answer 10

• probability of base b in position i (b,i)
• pseudocount to correct for finite number of input sequences
• sigma to represent general probability of base occurrence
• score sites to indicate certainty/uncertainty of particular base at that position
• search sequence for objects that are likely to arise from that PWM
• score directly related to binding energy of DNA-protein interaction
  
  • statistical and energy-based model

Question 11

assumptions of DNA motif knowledge approach

Answer 11

• nucleotide at one position has no effect on nucleotide present at adjoining position
• TFs have strict spatial requirements in binding sites that preclude variable spacing

Question 12

de novo prediction of DNA binding sites

Answer 12

• use of gene expression studies to identify coexpressed genes
• something upstream of coexpressed genes may explain expression behaviour
• statistical methods to identify the motif of interest in available sequences