Pathline: " 1234B2# ="3;6<89>";>6"B A1"<6"A <?;>"=21 9>";52#A. Miriah Meyer 1,2

Transcription

1 " 1234B2# ="3;6<89>";>6"B A1"<6"A <?;>"=21 9>";52#A +$ +" +% 1"!"#!"#"$!$!)!" / 0!"*!%%!" #" #$ #%!$ #&!% #' #( #), Pathline: - -*!". -+!". A Tool For Comparative -, Functional Genomics #%. 2.3'. "3.. Miriah Meyer 1,2 #$( Bang Wong 2 Mark Styczynski 3 Tamara Munzner Hanspeter Pfister 12.3'. Harvard University "3.. 2 Broad Institute 3 Georgia Institute of Technology 4567/ :8 0;-;68,<;5=/>!$* %"&'()*+&"$ %"&,+-$!$ 4 University +8/6:5=?@0#65@A" of British Columbia +8/6:5=?@0#65@A$ +8/6:5=BCC 4 1 -% -' -( -) -. -/ -%0 -%% -%' -%( -%) $% E&3. 2D3% &% E*3% 2'3$ &' E&3) 2)3" &( E)3' 2&3& $' E$3% 2*3. &) E$3* 2'3. $( E$3% 2*3. $( E$3% 2*3. &* E$3( 2%3$ &+ E&3% 2&3% &, ED3' 2$3% &'+ E"3) 2&3. E)3) 2*3D ED3% 2*3* E'3D 2*3( ED3' 2)3& E&3( 2(3( E(3' 2*3$ E)3* 2'3D E(3* 2*3" E*3' 2*3* E(3( 2D3% E'3' 2*3) E"3* 2%3. E"3( 2)3D ED3" 2*3. &!"%

2 roadmap background data & tasks Pathline case studies conclusions and future work

3 background

4 functional genomics how do genes work together to perform different functions in a cell?

5 functional genomics data gene expression molecular pathways

6 gene expression is the measured level of how much a gene is on or off... a single quantitative value 0.2

7 gene expression is the measured level of how much a gene is on or off... a single quantitative value biologists measure it for many genes g1 g2 g3 g g5-0.5 g6-0.7 g7-1.0 g8-0.5

8 gene expression is the measured level of how much a gene is on or off... a single quantitative value biologists measure it for many genes... in many samples (time points, tissue types, species) g1 g2 g3 g4 g5 s1 s2 s3 s4 s g g g

9 gene expression is the measured level of how much a gene is on or off... a single quantitative value biologists measure it for many genes... in many samples (time points, tissue types, species) g1 g2 g3 g4 g5 s1 s2 s3 s4 s visualized with heatmaps [Wilkinson09][Saldanha04][Seo02][Eisen98] [Gehlenborg10][Weinstein08] encode value with color g6 g7 g

10 gene expression is the measured level of how much a gene is on or off... a single quantitative value it is measured for many genes... in many samples (time points, tissue types, species) visualized with heatmaps [Wilkinson09] [Saldanha04] [Seo02] [Eisen98] [Gehlenborg10] [Weinstein08] encode value with color augmented with clustering [Eisen98]

11 the functioning of a cell is controlled by many interrelated chemical reactions performed by genes input output / input output genes genes = cell function

12

13 glycolysis tca cycle pathways

14 [Kim09] [Bourqui09] [Barsky08] [Letunic08] [Junker06] [Alm05] [Mlecnik05] [Shannon03]

15 [Kim09] [Bourqui09] [Barsky08] [Letunic08] [Junker06] [Alm05] [Mlecnik05] [Shannon03]

16 [Kim09] [Bourqui09] [Barsky08] [Letunic08] [Junker06] [Alm05] [Mlecnik05] [Shannon03] bioinformatics.cnmcresearch.org/geneshelf

17 functional genomics how do genes work together to perform different functions in a cell? comparative functional genomics how do the gene interactions vary across different species?

18 collaborators: Regev Lab at the Broad Institute biology: metabolism in yeast data: multiple genes multiple time points multiple related species multiple pathways problem: existing tools can only look at a subset of this data comparative functional genomics how do the gene interactions vary across different species?

19 contributions Pathline first interactive tool for visualizing multiple genes, time points, species, and pathways linearized pathway representation for comparing quantitative data along a pathway curvemap visual encoding of temporal gene expression validation case studies describing efficiency gains and new biological findings

20 data & tasks

21 glycolysis metabolic pathways 10 to 50 pathways of interest inputs/outputs called metabolites tca cycle gene expression 6000 genes and 140 metabolites 6 time points 14 species of yeast s1 g1 m1 g2 m2 s2 s3 s4 t1 g1 t2 0.2 g2 t g1 0.2 g g t1 t2 t3 t4 t5 g1 0.2 g g g t1 t2 t3 t4 t5 t6 g1 0.2 g g g g t1 t2 t3 t4 t5 t6 g2 1.0 g g g g g t1 t2 t3 t4 t5 t6 g5-0.5g g g2 1.0 g g g6-0.7g g8-0.3 g3-0.7g g t t1 g1 t20 g7-1.0g g4 1.0 g g g5 g8-0.5g g s5 s6 g3 m3 g8-0.5g g g g similarity scores aggregate time series for a gene/metabolite over species g7-1.0g phylogeny g '"#()* '"#+%, '"#-./ '"#-./01 aggregate s1 s2 s3 t1 t2 t3 t4 t5 t6 g , t1 t2 t3 t4 t5 t6 g t1 t2 t3 t4 t5 t6... g ,, = 0.83 similarity of expression across species aggregate: Pearson, Spearman, others 2"#3$. '"#(.4 5"#&6$ 5"#7.$ 5"#$.( '"#,$0 8"#9.: 2"#.$-!"#$%& '"#;.& '"#&6+

22 tasks 1. study expression data as a time series look for peaks, valleys, time shifts 2. detailed comparison of a limited number of time series filter using pathways filter again using genes or species 3. comparison of similarity scores of genes along a pathway(s) 4. comparison of multiple similarity scores

23 Pathline

24 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ metabolic pathways

25 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ similarity scores

26 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ similarity scores

27 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ similarity scores

28 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ gene expression

29 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ phylogeny

30 Pathline design decisions

31 encode quantitative values with spatial position [Cleveland84] [Lam07]

32 encode quantitative values with spatial position [Cleveland84] [Lam07] instead of a.. topological layout linearized pathway

33 encode quantitative values with spatial position [Cleveland84] [Lam07] instead of a.. conditioned heatmap curvemap $% $& $'( $') $'& $'* $+, $+' $++ $+.'.+.-.%./.(.).& courtesy of M. Styczynski from JavaTreeview jtreeview.sourceforge.net/.*.',

34 Pathline linearized pathway representation

35 linearized pathway representation 0.0 pathway similarity score 1.0 common axes to compare similarity scores

36 linearized pathway representation pathway similarity score common axes to compare similarity scores bars and circles

37 linearized pathway representation pathway similarity score common axes to compare similarity scores bars and circles

38 linearized pathway representation pathway similarity score common axes to compare similarity scores bars and circles visual layer for attenuation color-code gene direction

39 linearized pathway representation 0.0 pathway similarity score 1.0 common axes to compare similarity scores bars and circles visual layer for attenuation color-code gene direction multiple similarity scores

40 3,4", &4,, )& %& linearized pathway %* 0!"!#!$ representation 2& 1 %$. / %&, common axes to compare similarity scores %&, bars and circles visual layer for attenuation color-code gene direction )* %&'!+, %&' %&( multiple similarity scores multiple pathways %*' )+ %* %++

41 unroll from pathway to ordered list of nodes /*0#12)30* +*$*,-.!"#$%& 423$0 6 6!"#$%& 7 7 '(%)* unroll and cut reinsert,#.,1.,%.,5. stylized marks and shared coordinate frame

42 " 1234B2# ="3;6<89>";>6"B 3,4", &4,, linearized pathway )& %* %& representation 2& 0 1!"!#!$ %$. / %&, putting it together... %&, use spatial position for similarity scores instead of topology )* %&' %&'!+, %&( topology is secondary %*' )+ %* %++

43 Pathline curvemap

44 curvemap &'(%!"#$!"#% '% $%!"# $ % $ = % ( = >? )*+, inspired by heatmaps base visual unit is a curve )'-. '& '( )&/ ') 012- '* )3-4 '+ #562 ',

45 curvemap &'(%!"#$!"#% '% $%!"# $ % $ = % ( = >? )*+, inspired by heatmaps base visual unit is a curve )'-. filled, framed line charts to enhance shape perception )&/ '& '( ') 012- '* )3-4 '+ #562 ',

46 curvemap &'(%!"#$!"#% '% $%!"# $ % $ = % ( = >? )*+, inspired by heatmaps base visual unit is a curve )'-. filled, framed line charts to enhance shape perception rows are species )&/ columns are genes/metabolites 012- '& '( ') '* )3-4 '+ #562 ',

47 curvemap A1"<6"A <?;>"=21 9>";52#A $% &% &' (% (' () C,1, 0%1E C%1' 0%1% C21) 0%1F inspired by heatmaps base visual unit is a curve filled, framed line charts to enhance shape perception rows are species columns are genes/metabolites overlays to enhance trend perception (* (+ (, (- (. (/ (%0 (%% (%' (%) (%* C'1( 0,1) C)1F 0F1F C)1$ 0%1$ C)1( 021E C'1E 0%1" C)1, 0%1% C'1, 0E1' C'1E 0%1, C"1F 0$1F C"1F 0,1E C"1E 0%1( C)1( C%1' C)1, 0E1' 021$ 0,1"

48 !"!#!$%!$&!$#!$'!()!($!((!(* +$ +$#,$ +$#,( +( +**,* +$* +*)," -. +& +$) / 0 +( +$) 1$!*) +(# )2)) $2)) )2)) $2))!"#!"#"$ 3456/ :.:57-;:4</= %"&'()*+&"$ %"&,+-$,7/594<>?? B(2# B*2' B*2* B&2A BA2) BA2) B"2* B*2# B#2$ B#2$.'.+.-.%./.(.).&.*.',.''.'+.'-.'% " :7;8<01:125="79>6; ?"68<5;@ 0123A1# ;>:="<10 8=":41#@ Demo

49 case studies

50 " 0123A1# <"2:5;78=":=5"A ;>:="<10 &# &" $" $"% %"&'()*+&"$ $# &' &) /"!"#!"#"$ $# $, * + -. $"(!'( 0(12( "1(( $"( 0(12( "1(( $'' $"'!"!# :+:57*;:4<-= $#% %"&,+-$ &7-594<>?.!54?@" &7-594<>?.!54?@# &7-594<ABB '% '& '( ') '* '+ ', '- '. '%/ '%% '%& '%( '%) $% C%1' 021# $& C)1, 0,1" C,1, 0,1" C%1' 021( C21) 0'1% C'1( 021# C)1D 0)1) C)1# C(1( C#1' 0'1) C'1E 0"12 C)1, 0#1" C'1, 0)1D C'1E 0'1D C(12 0#1D C(1( 0"1" C"1' 0"1( whole genome duplication both genes one gene

51 " 0123A1# <"2:5;78=":=5"A ;>:="<10 &# &" $" $"% %"&'()*+&"$ $# &' &) /"!"#!"#"$ $# $, * + -. $"(!'( 0(12( "1(( $"( 0(12( "1(( $'' $"'!"!# :+:57*;:4<-= $#% %"&,+-$ &7-594<>?.!54?@" &7-594<>?.!54?@# &7-594<ABB '% '& '( ') '* '+ ', '- '. '%/ '%% '%& '%( '%) $% C%1' 021# $& C)1, 0,1" C,1, 0,1" C%1' 021( C21) 0'1% C'1( 021# C)1D 0)1) C)1# C(1( C#1' 0'1) C'1E 0"12 C)1, 0#1" C'1, 0)1D C'1E 0'1D C(12 0#1D C(1( 0"1" C"1' 0"1( whole genome duplication both genes one gene

52 " 0123A1# <"2:5;78=":=5"A ;>:="<10 &# &" $" $"% %"&'()*+&"$ $# &' &) /"!"#!"#"$ $# $, * + -. $"(!'( 0(12( "1(( $"( 0(12( "1(( $'' $"'!"!# :+:57*;:4<-= $#% %"&,+-$ &7-594<>?.!54?@" &7-594<>?.!54?@# &7-594<ABB '% '& '( ') '* '+ ', '- '. '%/ '%% '%& '%( '%) $% C%1' 021# $& C)1, 0,1" C,1, 0,1" C%1' 021( C21) 0'1% C'1( 021# C)1D 0)1) C)1# C(1( C#1' 0'1) C'1E 0"12 C)1, 0#1" C'1, 0)1D C'1E 0'1D C(12 0#1D C(1( 0"1" C"1' 0"1( whole genome duplication both genes one gene

53 " 0123A1# <"2:5;78=":=5"A ;>:="<10 &# &" $" $"% %"&'()*+&"$ $# &' &) /"!"#!"#"$ $# $, * + -. $"(!'( 0(12( "1(( $"( 0(12( "1(( $'' $"'!"!# :+:57*;:4<-= $#% %"&,+-$ &7-594<>?.!54?@" &7-594<>?.!54?@# &7-594<ABB '% '& '( ') '* '+ ', '- '. '%/ '%% '%& '%( '%) $% C%1' 021# $& C)1, 0,1" C,1, 0,1" C%1' 021( C21) 0'1% C'1( 021# C)1D 0)1) C)1# C(1( C#1' 0'1) C'1E 0"12 C)1, 0#1" C'1, 0)1D C'1E 0'1D C(12 0#1D C(1( 0"1" C"1' 0"1( whole genome duplication both genes one gene

54 gene-level relationships " 0123A1# <"2:5;78=":=5"A +2++ &2++ (& %& %) /!"!#!$ 1& 0 %$ -. %&+ %&+

55 gene-level relationships $% 2*3) D)3# $& 2,3* D,3* $' 2)3* DF3" 2)3" D$3" 2&3* DF3* 2*3) D)3) 2&3* DF3" 2)3, D*3+ 2+3, D)3, 2)3' D$3' 2*3) D#3+ 2&3, D'3" 2*3& D#3# 2,3* D"3" 2)3' D&3' 2&3* D+3+ 2&3& DF3& () (* (+ (, (% (& (' (- (. ()/ ()) ()* ()+ (), ;>:="<10 %& (& -. %$ %&+ / 0 %) %&+ 1& +2++ &2++ $% $& $' () (* (+ (, (% (& " 0123A1# ;>:="<10

56 gene-level relationships $% 2*3) D)3# $& 2,3* D,3* $' 2)3* DF3" 2)3" D$3" 2&3* DF3* 2*3) D)3) 2&3* DF3" 2)3, D*3+ 2+3, D)3, 2)3' D$3' 2*3) D#3+ 2&3, D'3" 2*3& D#3# 2,3* D"3" 2)3' D&3' 2&3* D+3+ 2&3& DF3& () (* (+ (, (% (& (' (- (. ()/ ()) ()* ()+ (), ;>:="<10 %& (& -. %$ %&+ / 0 %) %&+ 1& +2++ &2++ $% $& $' () (* (+ (, (% (& " 0123A1# ;>:="<10

57 gene-level relationships $% 2*3) D)3# $& 2,3* D,3* $' 2)3* DF3" 2)3" D$3" 2&3* DF3* 2*3) D)3) 2&3* DF3" 2)3, D*3+ 2+3, D)3, 2)3' D$3' 2*3) D#3+ 2&3, D'3" 2*3& D#3# 2,3* D"3" 2)3' D&3' 2&3* D+3+ 2&3& DF3& () (* (+ (, (% (& (' (- (. ()/ ()) ()* ()+ (), ;>:="<10 %& (& -. %$ %&+ / 0 %) %&+ 1& +2++ &2++ $% $& $' () (* (+ (, (% (& " 0123A1# ;>:="<10

58 gene-level relationships $% 2*3) D)3# $& 2,3* D,3* $' 2)3* DF3" 2)3" D$3" 2&3* DF3* 2*3) D)3) 2&3* DF3" 2)3, D*3+ 2+3, D)3, 2)3' D$3' 2*3) D#3+ 2&3, D'3" 2*3& D#3# 2,3* D"3" 2)3' D&3' 2&3* D+3+ 2&3& DF3& () (* (+ (, (% (& (' (- (. ()/ ()) ()* ()+ (), ;>:="<10 %& (& -. %$ %&+ / 0 %) %&+ 1& +2++ &2++ $% $& $' () (* (+ (, (% (& " 0123A1# ;>:="<10

59 gene-level relationships $% 2*3) D)3# $& 2,3* D,3* $' 2)3* DF3" 2)3" D$3" 2&3* DF3* 2*3) D)3) 2&3* DF3" 2)3, D*3+ 2+3, D)3, 2)3' D$3' 2*3) D#3+ 2&3, D'3" 2*3& D#3# 2,3* D"3" 2)3' D&3' 2&3* D+3+ 2&3& DF3& () (* (+ (, (% (& (' (- (. ()/ ()) ()* ()+ (), ;>:="<10 %& (& -. %$ %&+ / 0 %) %&+ 1& +2++ &2++ $% $& $' () (* (+ (, (% (& " 0123A1# ;>:="<10

60 conclusions and future work

61 conclusions Pathline: first interactive tool for comparative functional genomics multiple: genes, time points, species, and pathways two new visual encodings curvemap for expression data with multiple dimensions linearized pathway representation for comparing quantitative data case studies: efficiency gains and new discoveries

62 future work automate pathway selection and linearization continue with Regev Lab: more data types apply ideas to other biological systems Pathline linearized pathway representation curvemaps beyond biology: curvemap vs heatmap

63 " 1234B2# ="3;6<89>";>6"B A1"<6"A <?;>"=21 9>";52#A +$ +" +% 1"!"#!"#"$ 2.3'. "3..!"!" -% #" #$!$ #%!$ / #& -'!% #' #( #) -( 0 -)!) questions?, - -*!".!".!"* #%.!$* %"&'()*+&"$ 2.3'. "3..!$!%% #$( 4567/ :8 0;-;68,<;5=/> %"&,+-$ +8/6:5=?@0#65@A" +8/6:5=?@0#65@A$ +8/6:5=BCC -+ -, -. -/ -%0 -%% -%' -%( -%) $% E&3. 2D3% &% 2'3$ &' 2)3" &( 2&3& $' 2*3. &) 2'3. $( 2*3. $( acknowledgements Regev lab: Michelle Chan, Courtney French, Jay Konieczka, Jenna Pfiffner, Aviv Regev, and Dawn Thompson E"3* Data Visualization Initiative at the Broad Institute: National E*3% E&3) Science E)3' E$3% Foundation E$3* E$3% under E$3% E$3( Grant E&3% ED3' E"3) to the Computing Research Association for the CIFellows Project 2*3. &* 2%3$ &+ 2&3% &, 2$3% &'+ 2&3. E)3) 2*3D ED3% 2*3* E'3D 2*3( ED3' 2)3& E&3( 2(3( E(3' 2*3$ E)3* 2'3D E(3* 2*3" E*3' 2*3* E(3( 2D3% E'3' 2*3) 2%3. E"3( 2)3D ED3" 2*3. &!"%