1
Workflow Figure 1: The workflow of the training process of our molecular generator under the reinforcement learning framework. During this loop, each token in the SMILES sequence might be also generated by the pre-trained model with a small probability, which was set manually as exploring rate. De novo drug molecule generation with generative adversarial networks X.Liu 1 , K.Ye 2 , H.W.T.van Vlijmen 1,3 , A.P.IJzerman 1 , G.J.P.van Westen 1 1 Drug Discovery and Safety, Leiden Academic Center for Drug Research, Einsteinweg 55, Leiden, The Netherlands, 2 Omics and Omics informatics, Xi’an Jiaotong University, 28 Xianning W Rd, Xi’an, China, 3 Janssen Pharmaceutica NV, Beerse, Belgium Introduction Quantitative structure-activity relationship (QSAR) modeling is important for drug discovery and well established [1] . Recently, deep learning has been introduced as a new method for the construction of QSAR models [2] . Now deep learning can also be applied to generate molecules with a desired activity, or inverse QSAR [3,4] . Aim Although several groups investigated the performance of deep learning for de novo drug design [3,4] , the diversity of the generated molecules was unsatisfactory [5] . Here, we propose a method that integrates an exploring strategy into generative adversarial networks to generate ligands for the A 2A adenosine receptor as various as possible. Conclusions 1. Our exploring strategy decreased the number of duplicate molecules. 2. A higher exploring rate could increase the diversity of the generated molecules, but it also output more undesired molecules. 3. The baseline helps the model generate more desired molecules and increase their scores. But this effect at the lower exploring rate was not as significant as at the higher exploring rate. Contact Information Xuhan Liu, MSc Email: [email protected] Phone: +31 71 527 4332 References 1. Cherkasov, A., et al., QSAR modeling: where have you been? Where are you going to? J Med Chem, 2014. 57(12): p. 4977-5010. 2. Lenselink, E.B., et al., Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set. J Cheminform, 2017. 9(1): p. 45. 3. Olivecrona, M., et al., Molecular de-novo design through deep reinforcement learning. J Cheminform, 2017. 9(1): p. 48. 4. Benjamin, S.-L., et al., Optimizing distributions over molecular space. An Objective-Reinforced Generative Adversarial Network for Inverse-design Chemistry (ORGANIC). 2017. 5. Benhenda, M. ChemGAN challenge for drug discovery: can AI reproduce natural chemical diversity? ArXiv e-prints, 2017. 1708. Performance Figure 2: (a) The average score at each epoch during the training process with different parameters. (b) Receiver operator characteristic (ROC) curve for the predictor with different machine learning methods. (c, d) Scatter plot showing the logP ~ weight distribution (c) and the principle components analysis (d) for predicted active molecules generated by our proposed model and REINVENT [3] through 10,000 times sampling, compared with real ligands of the A 2A adenosine receptor. Exploring rate=0.01 Exploring rate=0.1 REINVENT Pre- trained Baseline 0.0 0.1 0.0 0.1 -- -- Valid SMILES 99.32% 99.55% 95.85% 98.75% 99.38% 93.88% Desired SMILES 98.66% 99.23% 74.60% 80.09% 98.70% 0.70% Unique SMILES 98.26% 96.86% 73.02% 80.03% 97.51% 0.70% Diversity 0.7946 0.7866 0.8452 0.8510 0.8090 0.8724 Loss Function () = ( 1: − ) =1 log ( | 1:−1 ) Molecule Diversity [5] () = 1 || 2 , ∈(,) 1− | | | | (b) (a) Figure 3: The generated molecules by our proposed method (a) and REINVENT (c) and known ligands of A 2A adenosine receptor (b) in 5 clusters based on principle component analysis. (d) (c) (1) (3) (2) (4) (5) (b) (a) (c)

De novo drug molecule generation with generative ... · Workflow Figure 1: The workflow of the training process of our molecular generator under the reinforcement learning framework

Embed Size (px)

Citation preview

Page 1: De novo drug molecule generation with generative ... · Workflow Figure 1: The workflow of the training process of our molecular generator under the reinforcement learning framework

Workflow

Figure 1: The workflow of the training process of our molecular generator underthe reinforcement learning framework. During this loop, each token in the SMILESsequence might be also generated by the pre-trained model with a smallprobability, which was set manually as exploring rate.

De novo drug molecule generation with generative

adversarial networks

X.Liu1, K.Ye2, H.W.T.van Vlijmen1,3, A.P.IJzerman1, G.J.P.van Westen11 Drug Discovery and Safety, Leiden Academic Center for Drug Research, Einsteinweg 55, Leiden, The Netherlands, 2 Omics and Omics informatics, Xi’an Jiaotong University, 28 Xianning W Rd, Xi’an, China,

3 Janssen Pharmaceutica NV, Beerse, Belgium

Introduction• Quantitative structure-activity relationship (QSAR) modeling is

important for drug discovery and well established[1] .• Recently, deep learning has been introduced as a new

method for the construction of QSAR models[2].• Now deep learning can also be applied to generate

molecules with a desired activity, or inverse QSAR[3,4].

AimAlthough several groups investigated the performance ofdeep learning for de novo drug design[3,4], the diversity ofthe generated molecules was unsatisfactory[5]. Here, wepropose a method that integrates an exploring strategy intogenerative adversarial networks to generate ligands for theA2A adenosine receptor as various as possible.

Conclusions1. Our exploring strategy decreased the number of duplicate

molecules.2. A higher exploring rate could increase the diversity of the

generated molecules, but it also output more undesiredmolecules.

3. The baseline helps the model generate more desired moleculesand increase their scores. But this effect at the lower exploringrate was not as significant as at the higher exploring rate.

Contact InformationXuhan Liu, MSc

Email: [email protected]: +31 71 527 4332

References1. Cherkasov, A., et al., QSAR modeling: where haveyou been? Where are you going to? J Med Chem,2014. 57(12): p. 4977-5010.2. Lenselink, E.B., et al., Beyond the hype: deepneural networks outperform established methodsusing a ChEMBL bioactivity benchmark set. JCheminform, 2017. 9(1): p. 45.3. Olivecrona, M., et al., Molecular de-novo designthrough deep reinforcement learning. JCheminform, 2017. 9(1): p. 48.4. Benjamin, S.-L., et al., Optimizing distributionsover molecular space. An Objective-ReinforcedGenerative Adversarial Network for Inverse-designChemistry (ORGANIC). 2017.5. Benhenda, M. ChemGAN challenge for drugdiscovery: can AI reproduce natural chemicaldiversity? ArXiv e-prints, 2017. 1708.

Performance

Figure 2: (a) The average score at each epoch during the training processwith different parameters. (b) Receiver operator characteristic (ROC) curvefor the predictor with different machine learning methods. (c, d) Scatter plotshowing the logP ~ weight distribution (c) and the principle componentsanalysis (d) for predicted active molecules generated by our proposed modeland REINVENT[3] through 10,000 times sampling, compared with real ligandsof the A2A adenosine receptor.

Exploring rate=0.01 Exploring rate=0.1 REINVENTPre-

trained

Baseline 0.0 0.1 0.0 0.1 -- --

Valid SMILES

99.32% 99.55% 95.85% 98.75% 99.38% 93.88%

DesiredSMILES

98.66% 99.23% 74.60% 80.09% 98.70% 0.70%

UniqueSMILES

98.26% 96.86% 73.02% 80.03% 97.51% 0.70%

Diversity 0.7946 0.7866 0.8452 0.8510 0.8090 0.8724

Loss Function

𝐿(𝜃) = (𝑅 𝑌1:𝑇 − 𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒)

𝑡=1

𝑇

log 𝑃( 𝑦𝑡 | 𝑌1:𝑡−1)

Molecule Diversity[5]

𝐷(𝑀) =1

|𝑀|2

𝑎,𝑏 ∈(𝑀,𝑀)

1 −|𝑓𝑝𝑎 ∩ 𝑓𝑝𝑏|

|𝑓𝑝𝑎 ∪ 𝑓𝑝𝑏|

(b)(a)

Figure 3: The generated molecules by our proposed method (a) and REINVENT(c) and known ligands of A2A adenosine receptor (b) in 5 clusters based onprinciple component analysis.

(d)(c)

(1)

(3)

(2)

(4)

(5)

(b)(a) (c)