Supplementary Information

Transcription

1 Supplementary Information 1

2 List of Figures 1 Models of circular chromosomes. 2 Distribution of distances between core genes in Escherichia coli K12, arc based model. 3 Distribution of distances between core genes in Escherichia coli K12, circular model. 4 Periodograms without detrend: theoretical distribution and arc based model distribution. 5 Periodograms without symmetrization (with detrend), arc based model. 6 Effects of symmetrization and detrend on the theoretical distribution. 7 Periodograms computed on models displaying a signal of 117kb period. 8 Curves (non smoothed) tracing chromosomal distribution of genes along the left and right arcs in E. coli K12 with a period of 33kb. Core genes and genes in COGs classes are considered. 9 Curves (non smoothed) tracing chromosomal distribution of genes along the left and right arcs in E. coli K12 with a period of 117kb. Core genes and genes in COGs classes are considered. 10 Distribution of genes in a randomly generated genome. 11 Curves (non smoothed) tracing chromosomal distribution of genes along the left and right arcs in E. coli K12 with a period of 33kb. Transcription factors and all genes in COG classes are considered. 12 Curves (smoothed and non smoothed) tracing chromosomal distribution in all E. coli K12 genes along th left and the right arcs with a period of 33kb. 13 Curves tracing the number of non-empty periodic windows along the E. coli K12 chromosome and the average number of genes in these windows for the left and right arcs. 14 Pearson correlation curve and macrodomains. 15 Plot of transcription data obtained under two conditions of response to stress and of the SCCI* curve. 16 Sectors and Pearson correlation curve computed between the SCCI* curve and the log-phase Expression* curve, the acid-shock Expression* curve and the heat-shock Expression* curve. 17 Statistical analysis of sectors for 1000 randomly generated genomes. 18 Distribution of genes in Baba s dataset. 19 Functional distribution of genes on the chromosomal spiral of period 28571nt. List of Tables 1 Distribution of genes, core genes and ribosomal proteins located in leading and lagging strands. 2 Characteristics of the Escherichia coli K12 organism and genome. 3 Periods computed for the three arc based models and for the circular model on core genes. 4 Periods computed for the three arc based models and for the circular model, taking into account that a core gene (possibly several) might occur in an operon. Case I. 5 Periods computed for the three arc based models and for the circular model, taking into account that a core gene (possibly several) might occur in an operon. Case II. 6 Periodograms comparison between the Escherichia coli K12 genome and randomly generated genomes. 7 Sectors positions along the E. coli K12 chromosome. 8 SCCI* peaks coverage with a period of 117kb along the chromosome. 9 Coverage of Expression* peaks for periods of 33kb and 117kb. 10 Periods obtained (with c = 3) on several groups of genes for the two chromosomal models. 11 Description of Baba s and Gerdes datasets of essential genes. 12 Pearson correlation coefficients between pairs of curves in Fig.5A in the article. 2

3 Figure 1: Top: A circular chromosome is ideally split into two halves whose extremes correspond to the replication origin (ORI) and the (ideally well determined) termination site of replication (TER). The four strands are grouped into leading and lagging strands following the 5-3 DNA directionality (as indicated by the arrows in the figure). Left and right chromosome arcs are unambiguously determined by the orientation of the strand that has been sequenced. By definition, a leading strand is a DNA strand that is replicated continuously. In DNA replication, the strand that is made in the 5 to 3 direction by continuous polymerization at the 3 growing tip. The lagging strand is at the opposite side of the replication fork from the leading strand. Bottom: the four models considered in the article with the strands used to compute distances between genes located on them. 3

4 800 Escherichia coli str. K-12 substr. MG1655 occurences e+06 2e+06 distances Figure 2: Distribution of distances, along the two arcs of the E. coli K12 MG1655 chromosome for the full arc based model, between core genes with symmetrization. Lines in the grid are spaced every 33566nt. No smoothing has been applied to the data. 4

5 Figure 3: Circular model: histogram (top corner) and periodogram computed on the distribution of minimal distances between pairs of core genes in the E. coli K12 chromosome. The E. coli K12 histogram is truncated on the y-axis with an effect on the central column; the maximum y-value is at 706. Compare it to Figure 2B in the article. 5

6 Figure 4: Periodograms realized without detrend on the theoretical distribution and on E. coli K12 distance distribution with the full arc based model. Notice that for the E. coli K12 distance distribution, the maximum peak remains at with the same spectral value. Compare it to Figure 2 in the article (the curves are essentially the same). 6

7 Figure 5: Periodogram realized without symmetrization (with detrend) on the full arc based model. Comparison with Figure 2B in the article: the maximum peak remains at but the spectral values decrease considerably. 7

8 A B C D Figure 6: Effects of symmetrization and detrend on the theoretical distribution. A: No symmetrization is applied, detrend is realized. The curve is non linear and this is due to the artificial step induced by the triangle shape in the FFT analysis. B: No detrend and no symmetrization operations are applied. The curve appears identical to the one in Figure 2A in the article, but zooming on the curve displays a sinusoidal shape as shown in C. C: two zooms on the curve in B, where a sinusoidal shape is present. D: two zooms on the curve (obtained without detrend) in SI Fig. 4A, absence of the sinusoidal shape. Notice that zooms on the curve in Figure 2 in the article look essentially the same as the ones reported here for SI Fig. 4. 8

9 Figure 7: Periodograms computed on models displaying a signal at 117kb. A: COGs classes ABJKL extended with rrnas were analyzed for the circular model. B and C: The signal at 117kb is not found for the full set of genes in COGs classes ABJKL but it is recovered for COGs classes ABJKL where only genes with a codon bias above the mean are considered (that is, a gene g is considered in the set only if SCCI(g) µ SCCI ). Both, the circular and the full arc based models display the 117kb signal. For the circular model, the spectrum associated to 117kb grows from (for biased genes in ABJKL) to (for ABJKL+rRNAs) and this is due to the precise periodic location of rrnas along the chromosome. D and E: Baba s dataset of essential genes displays the 117kb signal for both models. Notice that the signal at 117kb for Baba s has Z=4.57 while for ABJKL+rRNAs it has Z=6.87; Z-scores represent here a good manner to discriminate a peak from the mean behavior. Compare P-values in Table 10. 9

10 Figure 8: Curves (non smoothed) tracing chromosomal distributions of genes along the left and right arcs in E. coli K12 with a period of 33566nt. Genes belong to different groups: core genes (red curve), genes involved in metabolism (violet), in information storage and processing (green) and in cellular processing and signaling (blue). Violet, blue and green curves are constructed from genes classified in the COGs classes listed in the figure legend. To compare with the corresponding smoothed curves in Figure 5A in the article and with SI Figure 9. 10

11 Figure 9: Same as in SI Figure 8 but with a period of 117kb. This period was determined as statistically significant in [Wright et al. 2007], and as one can notice, a regularity in the distribution of genes is much less evident here. 11

12 number of genes in window Left dist from origin (mod 12938) in kb Right dist from origin (mod 12938) in kb periods (kb) spectrum Figure 10: Distribution of genes in a random genome. Top. Curves (non smoothed) tracing chromosomal distribution of genes tagged as core, on a period of 33566nt (left) and on a period of 12938nt (right) along the left and right arcs. The period of 12938nt is the main period obtained through Fourier analysis of this random genome. The random generation of genomes with tagged genes is explained in Methods. Compare the plots on the left to Fig. 4CD in the article, where the signal is much sharper: the total number of genes tagged as core in a given position varies within the interval [30, 90] for the left arc of the E. coli K12 genome (Fig. 4C) against [40, 70] for the randomly generated genome. The curve is defined by one main maxima for E. coli K12 and by several comparable local maxima for the random genome. Similarly for the right arc (Fig. 4D and also Fig. 5B). A noisy signal is easily recognized on the plots on the right with the absence of a main local maxima for the left arc. Bottom. Periodogram resulting from the Fourier analysis of the random genome. 12

13 Figure 11: Curves (non smoothed) tracing chromosomal distribution of E. coli K12 genes along the left and right arcs with a period of 33566nt. Top: distribution of all genes regulated by known transcription factors; bottom: distribution of all genes classified in COGs classes. We observe that for both groups, the left arc presents a periodic colocalization of genes in a subinterval between 10kb and 20kb from the chromosomal origin (mod 33566nt), with an extra, less pronounced, colocalization for genes regulated by known transcription factors at around 30kb. The right arc displays a much less sharp distribution. The list of genes regulated by known transcription factors (version 6.2) has been downloaded at 13

14 number of genes in window Left dist from origin (mod 33566) in kb number of genes in window Right dist from origin (mod 33566) in kb Figure 12: Curves (smoothed and non smoothed) tracing chromosomal distribution of all E. coli K12 genes along the left and right arcs with a period of 33566nt. Even if the signal is less sharp here than for core genes (see Figures 4CD in the article), there is a clear tendency of all E. coli K12 genes to colocalize in specific periodic subintervals. Namely, on the left arc we find a periodic gene colocalization between 10kb and 20kb from the chromosomal origin (mod 33566nt), and on the right arc we observe the same subinterval but also another, less pronounced, at around 30kb from the origin. 14

15 2.4 Left arc 2 Right arc mean number of genes in windows mean number of genes in windows dist from origin (mod 33566) in kb dist from origin (mod 33566) in kb 50 Left arc 40 Right arc Number of non empty windows Number of non empty windows dist from origin (mod 33566) in kb dist from origin (mod 33566) in kb Figure 13: Curves tracing the number of non-empty periodic windows along the chromosome (bottom; windows should contain some E. coli K12 core gene) and the average number of genes in these windows (top) for the left and right arcs. Each point of the curves corresponds to the counting done on periodic windows (of 5500nt) centered at this point along the 33kb period. The shape of the curves is similar to the ones in Fig.4CD and Fig.5B. This means that within chromosomal facets displaying a higher number of genes, we observe a larger number of periodic windows containing core genes and an increasing number of genes within these windows. 15

16 Figure 14: Pearson correlation curve and macrodomains. A: E. coli chromosome containing macrodomains as defined in [Esnault et. al, 2007]. Four macrodomains are defined: one around the oric (Ori, green), one around the dif (Ter, clear blue) and two flanking the Ter region, named Left (blue) and Right (red). The figure was redrawn from the one in [Esnault et. al, 2007] B: Pearson correlation curve shown in Figure 3B in the article. The chromosomal segments corresponding to the four macrodomains are colored as in A. Both extremes of the Left macrodomain and one extreme of the Right macrodomain coincide with the boundaries of chromosomal sectors. Namely, sector C starts at position nt and the Left macrodomain starts at nt; sector E starts at position nt and the Left macrodomain ends at position nt; secteur A (on the right arc) starts at position nt and the Right macrodomain ends at position nt. 16

17 0.4 Origin Left Arc Terminus Right Arc Origin 0.3 SCCI* under acid-shock* SCCI* under heat-shock* Clockwise position from origin (kb) 0 Figure 15: Transcriptomic data obtained under two conditions of response to stress (top: acid shock; bottom: heat shock) are plot (blue curves) against SCCI* values of E. coli K12 genes (red curve). Curves are smoothed. 17

18 Figure 16: Sectors computed on the Pearson correlation curve between the SCCI* curve and the log-phase Expression* curve (top), the acid-shock Expression* curve (middle) and the heat-shock Expression* curve (bottom) are reported. While log-phase and acid-shock conditions determine essentially the same chromosomal sectors (with the exception of A that splits in three large sectors under stress conditions), the heat-shock chromosomal sectors are clearly redefined, with much smaller sectors and a number of sectors (18) which is higher than it is expected in a randomly generated genome (this value is estimated to be 15 as indicated in Figure 17). Sectors B, C, D are the only ones that are preserved under heat-shock conditions. Colors correspond to macrodomains as in SI Fig. 14A. 18

19 Number of random genomes Number of sectors 150 genomes with at most 15 sectors genomes with at most 11 sectors genomes with at most 8 sectors Number of random genomes Percentage of sectors Figure 17: Statistical analysis of sectors for 1000 randomly generated genomes. For each random genome, sectors are computed from the Pearson correlation curve obtained as in Fig. 3B. Top: distribution of the number of sectors determined for 1000 random genomes, with mean and standard deviation Bottom: for each random genome, we compute the percentage of sector strips that cover more than 50% of the SCCI* peaks falling in the sector. Percentages are computed by taking into account only sectors containing SCCI* peaks. The distribution of percentages for random genomes containing at most 8 (blue), 11 (green) and 15 (red) sectors is plotted. (Notice that µ = 15, µ σ = 11 and µ 2σ = 8, where µ, σ are mean and standard deviation of the distribution.) No random genome with at most 8 sectors displays all sectors where strips cover > 50% of the SCCI* peaks, as for the E. coli K12 genome. For the blue histogram, the number of genomes for each bin is reported. 19

20 Figure 18: Distribution of genes in Baba s dataset. Baba s dataset is plotted on the intervals of 33566nt (left) and nt (right) for the left and right arcs, respectively. Observe that for period 117kb, there is a region highlighting a concentration of genes along both arcs. 20

21 Figure 19: Functional distribution of genes on the chromosomal spiral of period 28571nt. Distribution of genes belonging to the three main COGs functional groups is computed by counting the number of genes located on the chromosomal spiral of period 28571nt (starting from the origin O) and on successive windows of 5500nt along chromosomal arcs. The three functional groups are: information storage and processing (based on COGs classes ABJKL), cellular processing and signaling (DMNOPTUVWYZ) and metabolism (CEFGHIQ). Original curves have been smoothed with a Gaussian smoothing window with a sd of 2000nt. Original and smoothed curves are plot. The complementarity of the violet and blue curves can be easily observed. Green and blue curves follow the same tendency. 21

22 Escherichia coli str. K12 substr. MG1655 chromosome Lagging Leading Total Number of genes Number of core genes Top 100 core genes Core genes near to ORI Core genes near to TER Core ribosomal proteins near to ORI 3(3) 46(46) 49(49) Core ribosomal proteins near to TER 1(2) 4(7) 5(9) Table 1: Distribution of genes, core genes and ribosomal proteins located in leading and lagging strands. Number of genes, number of core genes and number of core ribosomal proteins located on the leading and on the lagging strands of Escherichia coli str. K12 substr. MG1655 genome. Proximity of core genes to ORI and TER is considered. For ribosomal proteins, the number in parenthesis describes the total number of ribosomal proteins. 22

23 Group King Size GC Gr Shape Arran Mot Oxy Habitat Temp Pathogenic γ-proteobacteria Bact Neg Rod S, P Yes Fac Host-ass Meso No Table 2: Characteristics of the E. coli K12 organism and genome. Group, Kingdom [Bacteria (Bact)], Size, GC, Gram stain [Negative (Neg)], Shape, Arrangement [Singles (S), Pairs(P)], Motility, Oxygen Req. [Facultative (Fac)], Habitat [Host-associated (Host-ass)], Temperature Range [Mesophilic (Meso)], Pathogenic. 23

24 Arc based models On S all On S leading On S lagging On S circ c = (P < 10 4 ; Z=7.81) 33333(P = ; Z=7.13) (4.39) (2.99) (4.67) (2.59) 31373(2.15) 17266(2.53) 80000(2.33) 33566(P < 10 4 ; Z=7.81) (4.39) 33566(P < 10 4 ; Z=7.81) (4.39) 65753(P = ; Z=4.97) (3.82) 33566(2.89) (2.8) 58537(2.77) (2.41) (2.37) 65753(P = ; Z=4.97) (3.82) c = (P = ; Z=7.13) (4.67) c = (P = ; Z=4.97) 33333(P = ; Z=7.13) (4.67) Circular model 33566(P < 10 4 ; Z=8.19) (5.5) 64865(2.15) 33566(P < 10 4 ; Z=8.19) (5.5) 33566(P < 10 4 ; Z=8.19) (5.5) Table 3: Periods computed for the three arc based models and for the circular model on core genes. Periods are computed with FFT analysis based on S all, S leading, S lagging and S circ defined on core genes. P -values and Z-scores are reported in parenthesis for the first peak and only Z-scores are reported for subsequent peaks. Lists are ordered from the highest peak to the lowest, and all peaks are greater than a fixed threshold depending on c (as described in Methods). These models do not take into account the organization of genes into operons. 24

25 Arc based models Circular model On S all On S leading On S lagging On S circ c = (P = ; (P = ; Z=7.12) 33333(P = ; Z=8.66) (P = ; Z=7.59) 33333(P = ; e Z=6.7) Z=3.65) (P = ; Z=2.75) (P = ; Z=7.59) 33333(P = ; Z=3.65) (P = ; Z=7.59) (P = ; Z=7.12) c = (P = ; Z=8.66) (P = ; Z=6.7) c = (P = ; Z=7.12) 33333(P = ; Z=8.66) (P = ; Z=6.7) Table 4: Periods computed for the three arc based models and for the circular model, taking into account that a core gene (possibly several) might occur in an operon. Case I. This model, compared to the arc based model considered in the article, discards from the set of core genes, those genes that belong to an operon and that are not the first occurring core gene in the operon (notice that the condition does not require the first occurring core gene to be the very first gene of the operon). Periods are computed with the FFT analysis on S all, S leading, S lagging and S circ. P -values and Z-scores are reported in parenthesis for the first peak and the peaks at 33kb. Lists are ordered from the highest peak to the lowest, and all peaks are greater than a fixed threshold depending on c (as described in Methods). 25

26 Arc based models Circular model On S all On S leading On S lagging On S circ c = (P = ; (P = ; Z=6.28) 33333(P = ; Z=8.39) (P = ; Z=7.71) 33333(P = ; Z=3.98) Z=6.56) (P = ; Z=3.71) (P = ; Z=7.71) 33333(P = ; Z=3.98) (P = ; Z=7.71) (P = ; Z=6.28) (P = ; Z=6.28) c = 3 c = (P = ; Z=8.39) (P = ; Z=6.56) (P = ; Z=3.71) 33333(P = ; Z=8.39) (P = ; Z=6.56) Table 5: Periods computed for the three arc based models and for the circular model, taking into account that a core gene (possibly several) might occur in an operon. Case II. The model considers those genes which are either the first gene in the operon (not necessarily core, but such that the operon contains at least one core gene) or a core gene that does not occur in an operon. Periods are computed by the FFT analysis on S all, S leading, S lagging and S circ. P -values and Z-scores are reported in parenthesis for the first peak and for the peaks at 33kb. 26

27 P -value #M r > µ r + 3σ r Arc based models S all c = 2 < c = 3 < c = 4 < S leading c = c = c = S lagging c = c = c = Circular model c = 2 < c = 3 < c = 4 < Table 6: Periodograms comparison between the E. coli K12 genome and randomly generated genomes. The subscript r refers to randomly generated genomes and the subscript g to the reference genome. The P -value counts the number of randomly generated genomes displaying a maximal peak which is greater than the maximal peak for the reference genome and divides it by 10000, that is the number of random generations. Condition #M r > µ r + 3σ r counts, for each periodogram generated from random genomes, the number of peaks which are greater than µ r + 3σ r ; the number in the table corresponds to peaks validating the inequality for random genomes. Notice that from the data in the Table we can conclude that the expected number of peaks which are > µ r + 3σ r in a random genome equals the number of peaks > µ + 3σ in E. coli K12 and that the peaks for random genomes have usually a smaller spectral value. 27

28 Sector sizes along the E.coli chromosome Sectors on left arc Sectors on right arc A B C D E F G A Start position End position Sector size (kb) Table 7: Sector positions along the E. coli chromosome. Sector positions are computed from the origin. Note that sector A overlaps the right and the left chromosomal arcs and its total size is 1634kb. 28

29 Sectors - left arc Sectors - right arc A B C D E F G A All peaks on the SCCI* curve - 117kb period # peaks in strip # total peaks Peaks on the SCCI* curve with values > µ + σ # peaks in strip # total peaks Table 8: SCCI* peaks coverage with a period of 117kb along the chromosome. Peaks in the SCCI* curve are organized in sectors (see SI Table 7). Strips accomodating most peaks in each sector are computed with a periodicity of 117kb and a size of 39kb (that is one third of the period). Compare to Table 1 in the article: we find a much better peak coverage with the 33kb period. 29

30 Sectors - left arc Sectors - right arc A B C D E F G A All peaks on the Expression* curve - 33kb period # peaks in strip # total peaks Peaks on the Expression* curve with values > µ + σ - 33kb period # peaks in strip # total peaks All peaks on the Expression* curve - 117kb period # peaks in strip # total peaks Peaks on the Expression* curve with values > µ + σ - 117kb period # peaks in strip # total peaks Table 9: Coverage of Expression* peaks for periods of 33kb and 117kb. Peaks in the Expression* curve are thought to be organized in sectors and strips accomodating most peaks in each sector are computed with a periodicity of 33566nt (with strip size of 11189nt) and of 117kb (with strip size of 39kb) (that is one third of the period). A better peak coverage is found with the 33kb period. 30

31 Sets of genes Arc based models Circular model All genes (4295) (P = ; Z=10.03) 3e+05 Core genes (563) 33566(P < 10 4 ; Z=7.81) Core genes + rrna genes (585) (P = ; Z=6.66) Wright s genes (2247) (P = ; Z=8.21) Core genes in Wright s set (452) 33566(P = ; Z=5.94) ABJKL genes in Wright s set (396) ABJKL genes in Wright s set with SCCI-value > µscci(274) COG classes CEFGHIQ: metabolism (1365) COG classes DMNOPTU- VWYZ: cellular processing and signalling (965) COG classes ABJKL: information storage and processing (681) On Sall On Sleading On Slagging (P = ; Z=5.5) (P < 10 4 ; Z=5.84) e (P = ; Z=10.38) (P = ; Z=6.33) (P = ; Z=5.82) ABJKL + rrna genes (703) (P = ; Z=6.87) ABJKL genes with SCCI-value > µscci (274) All COG classes confounded (3533) (P < 10 4 ; Z=6.75) (P = ; Z=9.25) Core genes in COGs (542) 33566(P = ; Z=7.9) Gerdes essential genes (602) 94118(P = ; Z=5.471) Baba s essential genes (300) (P < 10 4 ; Z=4.57) e+05(P = ; Z=5.06) (P = ; Z=4.97) (P = ; Z=4.21) (P = ; Z=6.1) (P < 10 4 ; Z=4.62) (P < 10 4 ; Z=5.7) (P < 10 4 ; Z=5.68) e (P = ; Z=8.38) (P = ; Z=4.35) (P = ; Z=7.22) 8e (P = ; Z=5.16) 8e (P < 10 4 ; Z=5.62) 8e (P = ; Z=4.83) (P = ; Z=4.22) (P = 0.003; Z=5.51) 8e+05 4e+05 9e+05(P < 10 4 ; Z=5.66) (P = ; Z=8.15) (P = ; Z=7.13) (P = 0.012; Z=7.13) (P = ; Z=8.42) (P = ; Z=6.07) (P = ; Z=8.09) (P = ; Z=5.91) (P = ; Z=8.73) (P = ; Z=5.71) (P = ; Z=3.84) (P = ; Z=3.84) (P = ; Z=6) (P = 0.012; Z=8.76) 2e (P = ; Z=5.54) (P = ; Z=6.76) (P = ; Z=4.94) (P = ; Z=3.68) (P < 10 4 ; Z=8.19) (P < 10 4 ; Z=7.97) (P = ; Z=6.93) (P < 10 4 ; Z=7.63) (P < 10 4 ; Z=5.57) (P < 10 4 ; Z=5.81) e (P = ; Z=5.63) (P = ; Z=6.86) (P = 0.001; Z=6.37) (P = ; Z=5.26) (P < 10 4 ; Z=6.23) e (P = 0.007; Z=10.52) 33566(P < 10 4 ; Z=8.78) (P = ; Z=6.33) (P < 10 4 ; Z=6.66) Table 10: Periods obtained with c = 3 on several groups of genes for the two kinds of chromosomal models. Periods are detected with the same methods and parameterization which was used in the article to analyse core genes. Periods which are integer multiples of 33kb are highlighted in bold and periods which are integer multiples of 117kb are underlined. For the first peak in a model, we report P -value and Z-score (computed on generated random genomes) of the peak. The number of genes in a dataset is reported in parenthesis, after the dataset name. 31

32 Baba Gerdes COGs ABJKL A B J K L CEFGHIQ C E F G H I Q DMNOPTUVWYZ D M N O P T U V W Y Z R S Wright (2247) Wright ABJKL Wright CEFGHIQ Wright DMNOPTUVWYZ Core genes Total number of genes Table 11: Table describing the content of Baba s and Gerdes datasets of essential genes. The two datasets are decomposed in COGs functional classes. The total number of genes contained in COGs classes is also given for comparison. The decomposition of Baba s and Gerdes datasets of essential genes in terms of core genes or genes collected in Wright s dataset is also reported. 32

33 Left arc CEFGHIQ DMNOPTUVWYZ Core genes ABJKL 0.58 (p < 2.2e 16 ) -0.53(p < 2.2e 16 ) 0.97(p < 2.2e 16 ) CEFGHIQ -0.19(p < 2.2e 16 ) 0.65(p < 2.2e 16 ) DMNOPTUVWYZ -0.63(p < 2.2e 16 ) Right arc CEFGHIQ DMNOPTUVWYZ Core genes ABJKL (p < 2.2e 16 ) 0.89(p < 2.2e 16 ) -0.40(p < 2.2e 16 ) CEFGHIQ -0.70(p < 2.2e 16 ) 0.89(p < 2.2e 16 ) DMNOPTUVWYZ -0.34(p < 2.2e 16 ) Table 12: Pearson correlation coefficients between pairs of curves in Fig. 5A in the article. The Pearson correlation coefficient and the associated P -value computed using Student s t-test, have been calculated for each pair of curves in Fig. 5A. Pairs are represented in a matrix. Two matrices describe the left arc (top) and the right arc (bottom). ABJKL, CEFGHIQ, DMNOPTUVWYZ and Core genes refer to the four curves drawn in Fig. 5A. 33