Limits...
Artificial neural network cascade identifies multi-P450 inhibitors in natural compounds.

Li Z, Li Y, Sun L, Tang Y, Liu L, Zhu W - PeerJ (2015)

Bottom Line: Thus, multi-P450 inhibition leads to greater drug-drug interaction risk than specific P450 inhibition.The results indicate significant positive correlation between the PIS values and the number of inhibited P450 isoforms (Spearman's ρ = 0.684, p < 0.0001).Furthermore, chemical similarity calculations suggested that the prevailing parent structures of natural multi-P450 inhibitors were alkaloids.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Pharmacy Administration, Harbin Medical University , Harbin , China.

ABSTRACT
Substantial evidence has shown that most exogenous substances are metabolized by multiple cytochrome P450 (P450) enzymes instead of by merely one P450 isoform. Thus, multi-P450 inhibition leads to greater drug-drug interaction risk than specific P450 inhibition. Herein, we innovatively established an artificial neural network cascade (NNC) model composed of 23 cascaded networks in a ladder-like framework to identify potential multi-P450 inhibitors among natural compounds by integrating 12 molecular descriptors into a P450 inhibition score (PIS). Experimental data reporting in vitro inhibition of five P450 isoforms (CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4) were obtained for 8,148 compounds from the Cytochrome P450 Inhibitors Database (CPID). The results indicate significant positive correlation between the PIS values and the number of inhibited P450 isoforms (Spearman's ρ = 0.684, p < 0.0001). Thus, a higher PIS indicates a greater possibility for a chemical to inhibit the enzyme activity of at least three P450 isoforms. Ten-fold cross-validation of the NNC model suggested an accuracy of 78.7% for identifying whether a compound is a multi-P450 inhibitor or not. Using our NNC model, 22.2% of the approximately 160,000 natural compounds in TCM Database@Taiwan were identified as potential multi-P450 inhibitors. Furthermore, chemical similarity calculations suggested that the prevailing parent structures of natural multi-P450 inhibitors were alkaloids. Our findings show that dissection of chemical structure contributes to confident identification of natural multi-P450 inhibitors and provides a feasible method for virtually evaluating multi-P450 inhibition risk for a known structure.

No MeSH data available.


Related in: MedlinePlus

Comparison of NNC model I and ANN model I.(A) Illustration of the framework of NNC model I. La ∾ Lh represent the laddersubmodels in which the corresponding molecular descriptors were imported; Pa ∾ Phare the integrated PIS parameters. For each submodel, the correlation coefficientsbetween the normalized number of inhibited P450 isoforms and P450 inhibition scoresof the compounds in the training set (RTr) and thetesting set (RTe) are shown. Spearman’s rho for thecorrelation between the PIS values and the normalized numbers of inhibited P450isoforms was also calculated for each integrated PIS (top). (B) Illustration of theframework of ANN model I. (C) The AUROCs are 0.876 and 0.862 for discriminationbetween P450 inhibitors (n = 1–5) and P450 non-inhibitors(n = 0) using NNC model I and ANN model I, respectively. (D) TheAUROCs are 0.860 and 0.836 for identification of non-multi-P450 inhibitors(n = 0–2) and multi-P450 inhibitors (n = 3–5)using the two models, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4696407&req=5

fig-2: Comparison of NNC model I and ANN model I.(A) Illustration of the framework of NNC model I. La ∾ Lh represent the laddersubmodels in which the corresponding molecular descriptors were imported; Pa ∾ Phare the integrated PIS parameters. For each submodel, the correlation coefficientsbetween the normalized number of inhibited P450 isoforms and P450 inhibition scoresof the compounds in the training set (RTr) and thetesting set (RTe) are shown. Spearman’s rho for thecorrelation between the PIS values and the normalized numbers of inhibited P450isoforms was also calculated for each integrated PIS (top). (B) Illustration of theframework of ANN model I. (C) The AUROCs are 0.876 and 0.862 for discriminationbetween P450 inhibitors (n = 1–5) and P450 non-inhibitors(n = 0) using NNC model I and ANN model I, respectively. (D) TheAUROCs are 0.860 and 0.836 for identification of non-multi-P450 inhibitors(n = 0–2) and multi-P450 inhibitors (n = 3–5)using the two models, respectively.

Mentions: Structure diversity was considered to group compounds used for model training andvalidation. To evaluate the NNC model architecture based on structure diversity, allsimilar compounds were classified to the training set, and partial dissimilar compoundswere classified to the validation set (Fig. 1).Our results indicate that the PIS of each molecular descriptor included in NNC model I wasonly weakly correlated with the normalized number of inhibited P450 isoforms, withSpearman’s rho values ranging from 0.413 to 0.620. However, ladder-like data integrationby NNC dramatically increased the correlation between chemical structure and multi-P450inhibition. We verified that the PIS values exported from the final ANN submodel weresignificantly positively corrected with the normalized number of inhibited P450 isoforms(Spearman’s rho = 0.713, p < 0.0001, Fig. 2A). In comparison, ANN model I using the same nine molecular descriptorsonly contributed a Spearman’s rho of 0.677 (Fig.2B). Consistent with this, ROC curve analysis indicated a significant increase inthe area under the ROC (AUROC) for identifying P450 inhibitors and multi-P450 inhibitorsusing NNC model I, compared with ANN model I (p < 0.0001, Figs. 2C and 2D, and Table S5).10.7717/peerj.1524/fig-2Figure 2Comparison of NNC model I and ANN model I.


Artificial neural network cascade identifies multi-P450 inhibitors in natural compounds.

Li Z, Li Y, Sun L, Tang Y, Liu L, Zhu W - PeerJ (2015)

Comparison of NNC model I and ANN model I.(A) Illustration of the framework of NNC model I. La ∾ Lh represent the laddersubmodels in which the corresponding molecular descriptors were imported; Pa ∾ Phare the integrated PIS parameters. For each submodel, the correlation coefficientsbetween the normalized number of inhibited P450 isoforms and P450 inhibition scoresof the compounds in the training set (RTr) and thetesting set (RTe) are shown. Spearman’s rho for thecorrelation between the PIS values and the normalized numbers of inhibited P450isoforms was also calculated for each integrated PIS (top). (B) Illustration of theframework of ANN model I. (C) The AUROCs are 0.876 and 0.862 for discriminationbetween P450 inhibitors (n = 1–5) and P450 non-inhibitors(n = 0) using NNC model I and ANN model I, respectively. (D) TheAUROCs are 0.860 and 0.836 for identification of non-multi-P450 inhibitors(n = 0–2) and multi-P450 inhibitors (n = 3–5)using the two models, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4696407&req=5

fig-2: Comparison of NNC model I and ANN model I.(A) Illustration of the framework of NNC model I. La ∾ Lh represent the laddersubmodels in which the corresponding molecular descriptors were imported; Pa ∾ Phare the integrated PIS parameters. For each submodel, the correlation coefficientsbetween the normalized number of inhibited P450 isoforms and P450 inhibition scoresof the compounds in the training set (RTr) and thetesting set (RTe) are shown. Spearman’s rho for thecorrelation between the PIS values and the normalized numbers of inhibited P450isoforms was also calculated for each integrated PIS (top). (B) Illustration of theframework of ANN model I. (C) The AUROCs are 0.876 and 0.862 for discriminationbetween P450 inhibitors (n = 1–5) and P450 non-inhibitors(n = 0) using NNC model I and ANN model I, respectively. (D) TheAUROCs are 0.860 and 0.836 for identification of non-multi-P450 inhibitors(n = 0–2) and multi-P450 inhibitors (n = 3–5)using the two models, respectively.
Mentions: Structure diversity was considered to group compounds used for model training andvalidation. To evaluate the NNC model architecture based on structure diversity, allsimilar compounds were classified to the training set, and partial dissimilar compoundswere classified to the validation set (Fig. 1).Our results indicate that the PIS of each molecular descriptor included in NNC model I wasonly weakly correlated with the normalized number of inhibited P450 isoforms, withSpearman’s rho values ranging from 0.413 to 0.620. However, ladder-like data integrationby NNC dramatically increased the correlation between chemical structure and multi-P450inhibition. We verified that the PIS values exported from the final ANN submodel weresignificantly positively corrected with the normalized number of inhibited P450 isoforms(Spearman’s rho = 0.713, p < 0.0001, Fig. 2A). In comparison, ANN model I using the same nine molecular descriptorsonly contributed a Spearman’s rho of 0.677 (Fig.2B). Consistent with this, ROC curve analysis indicated a significant increase inthe area under the ROC (AUROC) for identifying P450 inhibitors and multi-P450 inhibitorsusing NNC model I, compared with ANN model I (p < 0.0001, Figs. 2C and 2D, and Table S5).10.7717/peerj.1524/fig-2Figure 2Comparison of NNC model I and ANN model I.

Bottom Line: Thus, multi-P450 inhibition leads to greater drug-drug interaction risk than specific P450 inhibition.The results indicate significant positive correlation between the PIS values and the number of inhibited P450 isoforms (Spearman's ρ = 0.684, p < 0.0001).Furthermore, chemical similarity calculations suggested that the prevailing parent structures of natural multi-P450 inhibitors were alkaloids.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Pharmacy Administration, Harbin Medical University , Harbin , China.

ABSTRACT
Substantial evidence has shown that most exogenous substances are metabolized by multiple cytochrome P450 (P450) enzymes instead of by merely one P450 isoform. Thus, multi-P450 inhibition leads to greater drug-drug interaction risk than specific P450 inhibition. Herein, we innovatively established an artificial neural network cascade (NNC) model composed of 23 cascaded networks in a ladder-like framework to identify potential multi-P450 inhibitors among natural compounds by integrating 12 molecular descriptors into a P450 inhibition score (PIS). Experimental data reporting in vitro inhibition of five P450 isoforms (CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4) were obtained for 8,148 compounds from the Cytochrome P450 Inhibitors Database (CPID). The results indicate significant positive correlation between the PIS values and the number of inhibited P450 isoforms (Spearman's ρ = 0.684, p < 0.0001). Thus, a higher PIS indicates a greater possibility for a chemical to inhibit the enzyme activity of at least three P450 isoforms. Ten-fold cross-validation of the NNC model suggested an accuracy of 78.7% for identifying whether a compound is a multi-P450 inhibitor or not. Using our NNC model, 22.2% of the approximately 160,000 natural compounds in TCM Database@Taiwan were identified as potential multi-P450 inhibitors. Furthermore, chemical similarity calculations suggested that the prevailing parent structures of natural multi-P450 inhibitors were alkaloids. Our findings show that dissection of chemical structure contributes to confident identification of natural multi-P450 inhibitors and provides a feasible method for virtually evaluating multi-P450 inhibition risk for a known structure.

No MeSH data available.


Related in: MedlinePlus