Derivation of docking score threshold for AtPARP

the correct pose were examined. The results are shown in Table 3.14 and Figure 5.9. Since in PLANTS, all poses are ranked by TOTAL SCORE in ascending order, it was also investigated, whether the correct pose could be found in pose number 1, which corresponds to the most negative TOTAL SCORE.

Table 3.14: Characteristics of docking procedure

characteristic HsPARP1 HsPARP2 AtPARP2

Mean of correctly docked ligands 129 123 111

Mean number of correctly docked ligands in pose 1 102 80 92 Mean percentage of correctly docked ligands in pose 1 79.5 % 65.5 % 83.3 % Docking runs with 80 % > of correctly docked ligands 9 7 3 Docking runs with 95 % > of correctly docked ligands 5 4 0 Structures that could not be docked in any docking run 2 2 13 Structures docked correctly in all 10 docking runs 100 85 98

The mean number of structures that were docked correctly for HsPARP1 is higher than the mean number of correctly docked structures for AtPARP1 and AtPARP2. In contrast to that, the number of structures that could be docked correctly in the first pose (and with most negative docking score) is highest for AtPARP1.

The feature of docking procedure that was used for determination of a new docking score was the number of structures that could be docked correctly in each of thte 10 docking runs.

Table 3.14 shows that 100 of 142 structures were docked correctly in all 10 docking runs into HsPARP1, while 85 and 98 structures were docked into AtPARP1 and AtPARP2 in all docking runs, respectively. These structures could be docked with confidence into the corresponding active sites and therefore were used for further analysis (Figure 3.20). The docking scores for all inhibitors, depending on whether they were docked with confidence into HsPARP1 and AtPARP2, are shown in Figure 5.9.

Figure 3.20: Number of structures being docked into HsPARP1 and AtPARP Structures that could be docked confidently

into both HsPARP1 and AtPARP1

98 89

100 77 85 100

HsPARP1 AtPARP1 HsPARP1 AtPARP2

Structures that could be docked confidently into both HsPARP1 and AtPARP2

Structures that could be docked into active sites of HsPARP1 and AtPARP1/2

The intersection of the sets of structures that were docked in 10 docking runs correctly contain most information for further analysis. The intersecting sets of structures that could be docked confidently into HsPARP1 and either AtPARP1 or AtPARP2 contained 77 and 89 structures, respectively and is displayed in Figure 3.20.

Based on the 77 structures that were correctly docked into both active sites of HsPARP1 and AtPARP1 in all 10 docking runs, a new docking threshold was developed. The same procedure was performed for the 89 overlapping structures for HsPARP1 and AtPARP2. The differences of the mean docking scores were analysed with two-sided unpaired Student’s t-tests at significance level α=0.005 (2.7.3) which were performed for each of the 77 and 89 structures, respectively. The results of thes tests are shown color-coded in Figure 3.21, together with the step of final derivation of new AtPARP docking thresholds.

For the final derivation of a new docking threshold, only the ligands that fulfilled the prerequisites AND whose mean docking scores were significantly different were taken into account. For those remaining ligands, the median of docking score distributions of the differences were defined as the new docking threshold for AtPARP1/2. This is illustrated in Figure 3.21, where the left panel (A, C) represents the derivation of the new dockings threshold for AtPARP2, whereas the right panel (B, D) represents the derivation of the new docking threshold for AtPARP1, respectively. The upper barplots (A, B) represent the mean docking score differences of the remaining ligands. The distribution of those differences are represented in the lower histrograms (C, D). The medians of these histograms are represented by red lines, which implicate the median difference of docking scores.

The number of structures that were docked confidently into AtPARP1 and AtPARP2, it was investigated which had significantly different docking scores (by means of unpaired two-sided Student t-tests at significance level of α = 0.005. The results are diplayed in Table 3.15.

Table 3.15: AtPARP1 and AtPARP2 docking score differences

AtPARP1 AtPARP2

P<0.005 P>0.005 cond. on ∆ P<0.005 P>0.005 cond. on ∆

∆ > 0 01 02 03 20 12 32

∆ < 0 56 18 74 37 20 57

conditional on P 57 20 77 57 32 89

Number of structures docked into AtPARP1 and AtPARP2 with significantly different docking scores in comparison to HsPARP1. Results of corresponding Chi-Squared test are shown in 5.6.5

Figure 3.21: HsPARP1 and HsPARP2 docking scores of 142 HsPARP1 inhibitors

A and B: docking analysis of inhibitors docked into HsPARP1. C: docking analysis of inhibitors docked into HsPARP1 and AtPARP2.

By all of the 77 structures that were docked into the active site of AtPARP1, only 3 had a positive mean difference of docking scores in comparison to HsPARP1 mean docking scores.

Amidst the 74 structures with negative mean differences, 56 of them were significant at Difference of docking scores of HsPARP1 inhibitors

confidently docked into AtPARP1 and AtPARP2

HsPARP1 vs. AtPARP1 Analysis of docking results

HsPARP1 vs. AtPARP2

/ HsPARP1 andAtPARP2 docking scores difference significant*/not significant*

* at confidence level = 0.995

/ HsPARP1 andAtPARP1 docking scores difference significant*/not significant*

0 -10

10 5 -5 -15 -20 -25 -30

0 -10

Docking score differences

10 5 -5 -15 -20 -25 -30

frequency

0 5 10 15

frequency

0 5 10 15

0 -10

10 -20 -30

0 -10

10 -20 -30

Docking score difference distribution

Docking score differences

Docking score difference distribution

C D

B A

significance level of 0.005. Pearsons Chi-Sqared test was used to test for an association of docking scores and statistical significance. The test results are displayed in 5.6.5.

In both cases, there is no evidence against i_j, therefore the null hypothesis cannot be rejected, indicating that there is an association between the significant score differences and whether these differences are positive or negative. Although the test gives no hint about the type of association, based on the data it can be assumed that if a significant difference between the HsPARP1 and AtPARP1 (or AtPARP2) docking score exists, then it is likely that this difference is negative, e.g. the TOTAL SCORE for HsPARP1 is more negative than the TOTAL SCORE for AtPARP1 or AtPARP2.

To develop a new docking score threshold for AtPARP1 and AtPARP2 based on the mean differences of docking scores and the Pearson’s Chi-Squared Test, the medians of the 57 significant score differences between AtPARP1 and HsPARP2 and AtPARP2 and HsPARP1 were calculated (red lines in Figure 3.21). The median of the significant docking scores for AtPARP1 was –10.55 and –4.97 for AtPARP2. These scores represent the differences of docking scores for known human PARP1 inhibitors that are docked into AtPARP1/2 active sites. Since the medians of differences are derived from structures that could be docked with confidence into the corresponding active sites and account for significant difference in docking scores, they were used to obtain an adjusted docking threshold for AtPARP1 and AtPARP2, based on the threshold for HsPARP1. The new thresholds are calculated as:

Table 3.16: Derivation of new docking thresholds for AtPARP1 and AtPARP2

Threshold HsPARP1 (3.5.5)

Difference to HsPARP1

Threshold new

AtPARP1 -155.45 - -10.55 = -144.91

AtPARP2 -155.45 - -4.97 = -150.49

The adjusted docking thresholds for AtPARP1/2 are -144.91 and -150.49, respectively. These docking thresholds can be used to either select compounds from a vendor’s database that pass the docking threshold or classify selected compounds as potentially active or inactive based on their docking score.