Systematic Annotation

Transcription

1 4/5/2012

2 Review RTFM PNAS 95:14863

3 The Gene Ontology Three directed acyclic graphs (aspects): Biological Process Molecular Function Subcellular Component

4 The Gene Ontology

5 The Gene Ontology

6 The AmiGO browser

7 The Gene Ontology How might we annotate genes with GO terms? How do we calculate the significance of the GO terms associated with a particular group of genes?

8 Associating GO terms How might we annotate genes with GO terms?

9 Associating GO terms How might we annotate genes with GO terms? By sequence homology (e.g., BLAST) By domain homology (e.g., InterProScan) Mapping from an annotated relative (e.g., INPARANOID) Human curation of the literature (e.g., SGD)

10 Associating GO terms: Evidence codes Experimental EXP: Inferred from Experiment IDA: Inferred from Direct Assay IPI: Inferred from Physical Interaction IMP: Inferred from Mutant Phenotype IGI: Inferred from Genetic Interaction IEP: Inferred from Expression Pattern Computational Analysis ISS: Inferred from Sequence or Structural Similarity ISO: Inferred from Sequence Orthology ISA: Inferred from Sequence Alignment ISM: Inferred from Sequence Model IGC: Inferred from Genomic Context RCA: inferred from Reviewed Computational Analysis Author Statement TAS: Traceable Author Statement NAS: Non-traceable Author Statement Curator Statement Evidence Codes IC: Inferred by Curator ND: No biological Data available Automatically-assigned IEA: Inferred from Electronic Annotation Obsolete NR: Not Recorded

11 The Gene Ontology How might we annotate genes with GO terms? How do we calculate the significance of the GO terms associated with a particular group of genes?

12 Sampling with replacement: Mutagenesis How many transformants do we have to screen in order to cover a genome?

13 Sampling with replacement: Mutagenesis How many transformants do we have to screen in order to cover a genome? Probability that a transformant has (1) disrupted gene: p m Number of genes in organsim:

14 Sampling with replacement: Mutagenesis How many transformants do we have to screen in order to cover a genome? Probability that a transformant has (1) disrupted gene: p m Number of genes in organsim: Probability that a specific gene is disrupted in a specific transformant: p d = p m ( 1 ) = p m (1)

15 Sampling with replacement: Mutagenesis How many transformants do we have to screen in order to cover a genome? Probability that a transformant has (1) disrupted gene: p m Number of genes in organsim: Probability that a specific gene is disrupted in a specific transformant: p d = p m ( 1 Probability of not disrupting that gene: ) = p m (1) p u = 1 p m (2)

16 Sampling with replacement: Mutagenesis Probability of not disrupting that gene: p u = 1 p m (3)

17 Sampling with replacement: Mutagenesis Probability of not disrupting that gene: p u = 1 p m (3) The probability of not disrupting that gene n independent times is: ( p u,n = 1 p ) n m (4)

18 Sampling with replacement: Mutagenesis Probability of not disrupting that gene: p u = 1 p m (3) The probability of not disrupting that gene n independent times is: ( p u,n = 1 p ) n m (4) And the probability of disrupting that gene n independent times is: ( p d,n = 1 p u,n = 1 1 p ) n m (5)

19 Sampling with replacement: Mutagenesis Probability of not disrupting that gene: p u = 1 p m (3) The probability of not disrupting that gene n independent times is: ( p u,n = 1 p ) n m (4) And the probability of disrupting that gene n independent times is: ( p d,n = 1 p u,n = 1 1 p ) n m (5) This is also the expected genome coverage.

20 Sampling with replacement: Mutagenesis p_i or coverage n

21 Sampling with replacement: General Cases Calculating the probability of zero events was easy. ( p 0,n = 1 p ) n m (6)

22 Sampling with replacement: General Cases Calculating the probability of zero events was easy. ( p 0,n = 1 p ) n m (6) What about exactly k events?

23 Sampling with replacement: General Cases Calculating the probability of zero events was easy. ( p 0,n = 1 p ) n m (6) What about exactly k events? Binomial distribution: ( ) n p k,n = p k k m(1 p m ) n k (7)

24 Sampling with replacement: General Cases Calculating the probability of zero events was easy. ( p 0,n = 1 p ) n m (6) What about exactly k events? Binomial distribution: ( ) n p k,n = p k k m(1 p m ) n k (7) What if there is more than one type of event?

25 Sampling with replacement: General Cases Calculating the probability of zero events was easy. ( p 0,n = 1 p ) n m (6) What about exactly k events? Binomial distribution: ( ) n p k,n = p k k m(1 p m ) n k (7) What if there is more than one type of event? Multinomial distribution: p k1,k 2,...,n = n! ki! p k i i (8)

26 Sampling without replacement: GO Annotation The binomial distribution assumes that event probabilities are constant: ( ) n p k,n = p k k m(1 p m ) n k (9)

27 Sampling without replacement: GO Annotation The binomial distribution assumes that event probabilities are constant: ( ) n p k,n = p k k m(1 p m ) n k (9) What if there are m virulence factors in our genome, and every time we discover one it is magically removed from our library?

28 Sampling without replacement: GO Annotation The binomial distribution assumes that event probabilities are constant: ( ) n p k,n = p k k m(1 p m ) n k (9) What if there are m virulence factors in our genome, and every time we discover one it is magically removed from our library? Hypergeometric distribution: ( m )( N m ) k n k p k,m,n = ( N (10) n)

29 Sampling without replacement: GO Annotation The binomial distribution assumes that event probabilities are constant: ( ) n p k,n = p k k m(1 p m ) n k (9) What if there are m virulence factors in our genome, and every time we discover one it is magically removed from our library? Hypergeometric distribution: ( m )( N m ) k n k p k,m,n = ( N (10) n) More than one disjoint type of label: p k1,k 2,...,m 1,m 2,...,n = ( mi k i ) ( N n) (11)

30 Extracting gene lists from JavaTreeView

31 The SGD GO Slim Mapper

32 Multiple Hypothesis Testing

33 Alternatives to Hierarchical Clustering GORDER and pre-clustering by SOM

34 Alternatives to Hierarchical Clustering GORDER and pre-clustering by SOM Pre-calling number of clusters: k-means and k-medians

35 Alternatives to Hierarchical Clustering GORDER and pre-clustering by SOM Pre-calling number of clusters: k-means and k-medians Principal Component Analysis (PCA)

36 Homework Download PyMol