1. New Methods for Comparative genomics
-Mark Yandell HHMI/BDGP

2. New Methods for Comparative genomics
-Mark Yandell HHMI/BDGP
CGL: a software library for comparative genomics
Explore recent history (5-70 myr years)
Explore ancient history ( myr years)

3. CGL: A software library for comparative genomics
Part I:
CGL: A software library for comparative genomics

4. >264
MSFNNALSGVNAAQKDLNVTANNIANVNTTGFKESRAEFADVYANSIFVNAKTQVGNGVA TGAVAQQFHQGALQFTNNALDLSIQGNGFFVTSDGLTNLDRTFTRAGAFKLNENSYMVNN QGNYLQGYEINTDGTPKAVSINATKPIQIPDRAGEPKMTELVEASFNLSIESKTKPTSPA AFDPTNSATFAHSTSVTIYDSLGAPHVITKYFVRHEDPAAPGTPLTPGVKMTFTSGKLDP TLTVPVDPIKTVALGTTAGIINNGADPTQTLEIRLGDVTQYSSPFNVTKLTQDGATVGNL TKVEITPDGIVSATYSNATTLKVAMVALAKFANSQGLTQVGDTSWRQSLLSGDALPGTPN SGTLGSIKSSALEQSNVDLTSQLVNLITAQRNFQANSRSLEVNSSLQQTILQI
>264 ATGAAAGTTAGTTTTGAAAGAATAATTCCAAGTGAAAAAAGCTCTTTCCGCACACTGCAT AATAACTCTCCTATTTCTGAATTTAAATGGGAGTATCATTATCATCCGGAAATAGAACTG GTATGTGTAATTTCGGGAAGTGGCACACGGCATGTAGGCTACCATAAAAGCAATTATACA AACGGAGATCTTGTGTTAATAGGTTCAAACATTCCACATTCCGGATTTGGACTGAATTCT GTTGATCCGCATGAAGAAATAGTACTTCAGTTCAGGGAAGAGATTTTGCATTTTCCACAA CAGGAAGTTGAAACAAGAGCCGTGAAAGATCTACTGGAACGCTCTAAATATGGTATTCTG TATAGTACAGCTACAAAAAAGCTGCTCATGCCGAAACTAAAAAAGCTTCTGGAATCCGAA GGCTACAAAAGATACTTACTACTTCTGGAGATTCTCTTCGAACTTTCTTTGTGCGAGGAA TATGAATTGTTGAACAAAGAAATTATGCCTTATACCATAATCTCTAAAAATAAAACAAGA CTGGAAAATATCTTTACCTATGTGGAACATCATTACGATAAGGAAATAAATATAGAGGAT GTTGCAAAGCTGGCTAATCTTACTCTTCCTGCATTTTGTAATTTTTTTAAAAAAGCAACA CAGATTACCTTTACAGAATTTGTCAACCGTTACCGTATTAATAAAGCCTGCCTTCTGATG ACTCAGGATAAAACAATATCCGAATGCAGCTACAGTTGTGGCTTTAACAATGTTACTTAT TTCAACAGAATGTTTAAAAAATATACCAATAAAACGCCATCAGAATTT

6. Gene structure Sequence similarity ? ? ? …DEW RQWKMDKQV WDA ER
…DDW RGWKMDKQV WDMER ?
Gene structure
Sequence similarity

7. Aligning two sequences implicitly aligns two genes
gacgagtgg
gacgagtggcgacaatggaaaatggacaaacaagtg
tgggacgctgagcga
agtttaaagc
atgatatatatatatatatcg gt
ag gt
ag taa
…DEW
RQWKMDKQV
WDAER
…DDW
RGWKMDKQV
WDMER gt
ag gt
ag tga
agaaatttcg
atgcgcgcgcgcgcgcg
gatgattgg
gatgggtggcgcgggtggaaaatggacaaacaggtg
tgggacatggagcga

8. Using BLASTP alignments to make sense of genomic annotations
Annotation A
gacgagtgg
gacgagtggcgacaatggaaaatggacaaacaagtg
tgggacgctgagcga ?
agtttaaagc
atgatatatatatatatatcg gt
ag gt
ag taa
…DEW
RQWKMDKQV
WDAER
…DDW
RGWKMDKQV
WDMER gt
ag gt
ag tga
agaaatttcg
atgcgcgcgcgcgcgcg
gatgattgg
gatgggtggcgcgggtggaaaatggacaaacaggtg
tgggacatggagcga
Annotation B

9. Using genomic annotations to make sense of BLASTP alignments
Score = 42.0 bits (97), Expect(2) = 2e-25 Identities = 23/64 (35%), Positives = 40/64 (61%), Gaps = 2/64 (3%) Frame = -3 Query: 1 MFQNDVSSPRELQLMAAKVEKELGPVDILVNNASLMPMTSTP-SLKSDEIDTILQLNL-G DVS+ E++ M KV + GP+DIL+NNA ++ T P + +E D ++ +NL G Sbjct: 29 VVKADVSNREEVREMVKKVIDKFGPIDILINNAGILGKTKDPLEVTDEEWDRVISVNLKG Score = 37.7 bits (86), Expect(2) = 2e-25 Identities = 17/43 (39%), Positives = 29/43 (66%) Frame = -1 Query: 60 VTGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDI TGA G+GRAI++ELAK+G IN SGAE+ K+ +++ Sbjct: 89 ITGASRGIGRAIAIELAKRGVNVV---INYSGAEEEAKKTEEL 128

10. Using genomic annotations to make sense of BLASTP alignments
Score = 42.0 bits (97), Expect(2) = 2e-25 Identities = 23/64 (35%), Positives = 40/64 (61%), Gaps = 2/64 (3%) Frame = -3 Query: 1 MFQNDVSSPRELQLMAAKVEKELGPVDILVNNASLMPMTSTP-SLKSDEIDTILQLNL-G DVS+ E++ M KV + GP+DIL+NNA ++ T P + +E D ++ +NL G Sbjct: 29 VVKADVSNREEVREMVKKVIDKFGPIDILINNAGILGKTKDPLEVTDEEWDRVISVNLKG Score = 37.7 bits (86), Expect(2) = 2e-25 Identities = 17/43 (39%), Positives = 29/43 (66%) Frame = -1 Query: 60 VTGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDI TGA G+GRAI++ELAK+G IN SGAE+ K+ +++ Sbjct: 89 ITGASRGIGRAIAIELAKRGVNVV---INYSGAEEEAKKTEEL 128
splice junction
splice junction
splice junction
splice junction
splice junction
splice junction

11. ? Annotation A Some un-annotated genome …DEW gacgagtgg
gacgagtggcgacaatggaaaatggacaaacaagtg
tgggacgctgagcga ?
agtttaaagc
atgatatatatatatatatcg gt
ag gt
ag taa
…DEW
RQWKMDKQV
WDAER
Some un-annotated genome

12. Using genomic annotations to make sense of TBLASTX alignments
Score = 112 bits (240), Expect(2) = 2e-25 Identities = 58/99 (58%), Positives = 61/99 (61%) Frame = -3 / Query: VTSATLPLTSFSLSPSANR*SNRCLRKMASRTRGRRSRKTMTFRLKMARVVSSLRCACAA TSATLPLTSFSL P ANR*S RC MASR RG SRK + FRLK ARVVSSLR AC A Sbjct: LTSATLPLTSFSLRPRANR*SRRCFSPMASRIRGSSSRKVRIFRLKRARVVSSLRRACEA 45792
Query: ACVAVGVAEAVGPVTVT*PAGTAVVVRVMDSRKPAVSTT A VAVG * V+RVMDS+KPAVSTT Sbjct: AWVAVGAVGVAEDEGTV*AGPAGAVMRVMDSMKPAVSTT 45909

13. Using genomic annotations to make sense of TBLASTX alignments
Score = 112 bits (240), Expect(2) = 2e-25 Identities = 58/99 (58%), Positives = 61/99 (61%) Frame = -3 / Query: VTSATLPLTSFSLSPSANR*SNRCLRKMASRTRGRRSRKTMTFRLKMARVVSSLRCACAA TSATLPLTSFSL P ANR*S RC MASR RG SRK + FRLK ARVVSSLR AC A Sbjct: LTSATLPLTSFSLRPRANR*SRRCFSPMASRIRGSSSRKVRIFRLKRARVVSSLRRACEA 45792
Query: ACVAVGVAEAVGPVTVT*PAGTAVVVRVMDSRKPAVSTT A VAVG * V+RVMDS+KPAVSTT Sbjct: AWVAVGAVGVAEDEGTV*AGPAGAVMRVMDSMKPAVSTT 45909
1st Coding Exon
5’-UTR
1st Intron
2nd Coding Exon

14. Using genomic annotations to make sense of TBLASTX alignments
Score = 112 bits (240), Expect(2) = 2e-25 Identities = 58/99 (58%), Positives = 61/99 (61%) Frame = -3 / Query: VTSATLPLTSFSLSPSANR*SNRCLRKMASRTRGRRSRKTMTFRLKMARVVSSLRCACAA TSATLPLTSFSL P ANR*S RC MASR RG SRK + FRLK ARVVSSLR AC A Sbjct: LTSATLPLTSFSLRPRANR*SRRCFSPMASRIRGSSSRKVRIFRLKRARVVSSLRRACEA 45792
Query: ACVAVGVAEAVGPVTVT*PAGTAVVVRVMDSRKPAVSTT* A VAVG * V+RVMDS+KPAVSTT* Sbjct: AWVAVGAVGVAEDEGTV*AGPAGAVMRVMDSMKPAVSTT* 45909
1st Coding Exon
5’-UTR ?
1st Intron
2nd Coding Exon ?

15. Using TBLASTN alignments to identify orthologus exons and introns
in an un-annotated genome
Score = 53.5 bits (127), Expect(2) = 3e-10 Identities = 29/58 (50%), Positives = 30/58 (51%) Frame = -2 Query: 1 MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEVMVIT MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEV+V++ Sbjct: 2097 MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEVAVVS 2270
splice junction
1st coding exon
orthologus intron
begins here (2252)
Score = 105 bits (261), Expect(2) = 3e-10 Identities = 52/52 (100%), Positives = 52/52 (100%) Frame = -1 Query: 49 SIAGEVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK EVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK
Sbjct: 3991 NVMPEVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK 4146
splice junction
2nd coding exon
orthologus intron ends here (4003)

16. Using TBLASTN alignments to identify orthologus exons and introns
in an un-annotated genome
Score = 53.5 bits (127), Expect(2) = 3e-10 Identities = 29/58 (50%), Positives = 30/58 (51%) Frame = -2 Query: 1 MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEVMVIT MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEV+V++ Sbjct: 2097 MSEVVRNTLVKLQAIIALIGLAAITPLLILVALLGRLIAKLCWCSAPKSIAGEVAVVS 2270
splice junction
1st coding exon
‘1st coding exon’
orthologus intron
begins here (2252)
Score = 105 bits (261), Expect(2) = 3e-10 Identities = 52/52 (100%), Positives = 52/52 (100%) Frame = -1 Query: 49 SIAGEVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK EVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK
Sbjct: 3991 NVMPEVMVITGAGHGLGRAISLELAKKGCHIAVVDINVSGAEDTVKQIQDIYKVRAK 4146
splice junction
2nd coding exon
'2nd coding exon’
orthologus intron ends here (4003)

17. Exploring recent history with CGL
Part II
Exploring recent history with CGL
D. melanogaster Intron length distribution

18. time (millions of years)
D. virillis
D. pseudoobscura
D. ananassae
D. yakuba
D. simulans
D. melanogaster
62.9
54.9
44.2
12.8
5.4
time (millions of years)
Tamura et al (2003) Temporal Patterns of Fruit Fly (Drosophila Evolution Revealed
by Mutation Clocks ; MBE

19. Compare intron lengths
agtttaaagc
Query: 2 gacgagtgg gaccagtggcgacaatggaaaatggacaaacaagtg tgggacgctgagcga 54
|| || ||| || ||| ||||||||||||||||||||| |||||||||||||||
Sbjct: 29 gatgattgg gatgggtgg---gggtggaaaatggacaaacaggtg tgggacatggagcga 84
agaaatttcg
atgcgcgcgcgcgcgcg
atgatatattatatatatcg
DDWRGW-MDKQVWDMER
DEWRQWKMDKQVWDAER
Query: 2
Sbjct: 29 18 44
D+WR WKMDKQVWDMER
How do intron lengths change over time?
Compare intron lengths
Contig or read
MDSR
DKQV
MESR
atggatagtagc
gataagcagaag
atggaaagtagc
gataaacaaaag
annotated
inferred Li
Li’
Annotation

20. Compare intron lengths
agtttaaagc
atgatatattatatatatcg
Query: 2 gacgagtgg gaccagtggcgacaatggaaaatggacaaacaagtg tgggacgctgagcga 54
|| || ||| || ||| ||||||||||||||||||||| |||||||||||||||
Sbjct: 29 gatgattgg gatgggtgg---gggtggaaaatggacaaacaggtg tgggacatggagcga 84
agaaatttcg
atgcgcgcgcgcgcgcg
Compare intron lengths
Contig or read
MDSR
DKQV
MESR
atggatagtagc
gataagcagaag
atggaaagtagc
gataaacaaaag
annotated
inferred Li
Li’
Annotation
Log10 length D. simulans
Log10 length D. melanogaster

21. Why do intron lengths change over time?
D. pseudoobscura intron has transposon
D. melanogaster intron has transposon
Both introns have the same transposon
D. pseudoobscura intron length
D. melanogaster intron length

22. Does selection on the protein influence intron lengths ?
[(Li +Lj) - |Li–Lj| ]/Li+Lj
Annotation
gataagcagaag
atggatagtagc Ka Li
annotated
MDSR
DKQV Lj
MESR
DKQV
inferred
atggaaagtagc
gataaacaaaag
Contig or read

23. melanogaster & simulans melanogaster & yakuba
5 million yrs
13 million yrs
55 million yrs
250 million yrs
melanogaster & simulans
melanogaster & yakuba
melanogaster & pseudoobsura
melanogaster & mosquito
A new evolutionary clock?

24. A new evolutionary clock
[(Li +Lj) - |Li–Lj| ]/Li+Lj
D. simulans 5.4 myr
D. yakuba myr
D. pseudoobscura 54.9 myr
D. Virilis myr
D.ananassae myr

25. P(6) vs. dpse << 0.001 P(6) vs. dyak = 0.001 Region appears
To be un-duplicated
In dpse; might be duplicated in dyak
P(6) vs. dpse << 0.001
P(6) vs. dyak = 0.001

26. Are the phenoma we are seeing in th flies common to other orgs?
Explain the experiment
What does this tell us?
AA: recripcol best hits seems to work. Do it al the time but how do we know it workks: these data provide some independent confirnation of recip-best hits approachs.
A. most paralogs much older than 70 million years– current details as to genes and gene families established much, much more than 70 million years ago.
B. Paralogs diverging from one-another at same rate in both genomes– on average

27. Exploring ancient history with CGL
Part III
Exploring ancient history with CGL
C. elegans
C. intestinalis
H. sapiens
M. musculus
A.gambiae
D. melanogaster
1000
time (myr)
250 70

28. protein similarities
similarities in the intron-exon structures of genes

29. MIKEVFRPDKFGMDL
MILEVFRPDKFGMDL
MIKDVFRPDKFGIDL

30. Fraction of identical amino acids
The evolutionary trajectory of protein similarities
Best HSP, reciprocal best blastp hits
Human vs. mouse
Dmel vs. agam
Human vs. ciona
Human vs. plant
Fraction of identical amino acids

31. Fraction of identical amino acids
The evolutionary trajectory of protein similarities
Asymptotic distribution?
70 million years
250 million years
600 million years
3 billion years
Fraction of identical amino acids
time

32. Fraction of identical amino acids
D. melanogaster
vs.
A. gambiae
H. sapiens
vs.
C. intestinalis
Fraction of identical amino acids
C. intestinalis
H. sapiens
M. musculus
A.gambiae
D. melanogaster
C. elegans

33. Fraction of identical amino acids?
Fraction of identical amino acids

34. Organism A
Organism B
BLASTP
All proteins
All proteins
Every reciprocal best hit
For every HSP
Global similarity between the two organisms’ proteomes

35. H ‘standard’ C. elegans C. intestinalis H. sapiens M. musculus
A.gambiae
D. melanogaster
time (myr)
1000
250 70
C. elegans
C. intestinalis H
100
H. sapiens
100
M. musculus
A.gambiae
100
D. melanogaster

36. MIKEVFRPDKFGMDL
MILEVFRPDKFGMDL
MIKDVFRPDKFGIDL

37. I = * Organism A Organism B BLASTP All proteins All proteins
Every reciprocal best hit
XXXXXXXXXXXXXX
For every HSP
qintron-intron
I =
pquery
intron *
psbjct
intron
Global similarity between the two organisms’ intron-exon structures

38. Good trees can be made from both measures; these measures provide a rapid
and robust means to calculate phylogenetic trees using the combined
data from many annotated genomes simultaneously.
C. elegans
C. elegans
C. intestinalis
C. intestinalis
100
100
H. sapiens
H. sapiens
100
100
M. musculus
M. musculus
A. gambiae
A. gambiae
100
100
D. melanogaster
D. melanogaster H I
Trees based on I provide an independent check on trees made from
amino-acid similarities; note that deep nodes are better resolved in the
I tree.

39. Like intron lengths, gene-structures seem to evolve independently of protein
Sequence– hold H constant and you still get the same I tree
A. gambiae
D. melanogaster
A. gambiae
D. melanogaster
C. elegans
M. musculus
M. musculus
C. elegans
H. sapiens
H. sapiens
C. intestinalis
C. intestinalis I
H = 1.5
~ 50% similarity

40. New Methods for Comparative genomics
Part I: CGL: a software library for comparative genomics
Genome annotations make possible any new kinds of genome analyses
We have a developed a (soon to be) publicly available software library
called CGL; ‘seagull’ that that greatly facilitates such analyses.
Part II: Using CGL to explore recent history (1-70 myr years)
Intron length can be used as an evolutionary clock.
Seems to evolve independently of protein sequence.
Can be used to confirm and calibrate existing protein clock approaches.
Can be used to identify recent gene duplication events.
Can be used to strengthen and extend with many ‘standard’ protein based approaches, e.g. quartet analysis.
Part II: Using CGL to explore ancient history ( myr years)
Proteins may saturate completely after ~ 1 billion years or so.
Resolving deep evolutionary questions w/ proteins means computing ‘consensus’ trees; we offer one approach to this problem using ‘H’.
Like intron length, the intron-exon structures of genes contains a phylogenetic signal; the ‘I’ vs. ‘H’ trees
This signal seems to evolve independently of protein sequence, and perhaps more slowly.
Large scale comparative genomics has much to tell us about the evolution of both genes and organisms.

41. Acknowledgements Chris Mungall Simon Prochnik Chris Smith
Josh Kaminker
George Hartzell
Suzi Lewis
Gerald M. Rubin