Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pathway-Guided Optimization of Deep Generative Molecular Design Models for Cancer Therapy

Alif Bin Abdul Qayyum, Susan D. Mertins, Amanda K. Paulson, Nathan M. Urban, Byung-Jun Yoon

TL;DR

This work proposes how mechanistic models - such as pathway models described by differential equations - can be used for effective latent space optimization of JTVAEs and other similar models for GMD, and shows how a pharmacodynamic model can be incorporated for effective LSO of data-driven models for GMD.

Abstract

The data-driven drug design problem can be formulated as an optimization task of a potentially expensive black-box objective function over a huge high-dimensional and structured molecular space. The junction tree variational autoencoder (JTVAE) has been shown to be an efficient generative model that can be used for suggesting legitimate novel drug-like small molecules with improved properties. While the performance of the generative molecular design (GMD) scheme strongly depends on the initial training data, one can improve its sampling efficiency for suggesting better molecules with enhanced properties by optimizing the latent space. In this work, we propose how mechanistic models - such as pathway models described by differential equations - can be used for effective latent space optimization(LSO) of JTVAEs and other similar models for GMD. To demonstrate the potential of our proposed approach, we show how a pharmacodynamic model, assessing the therapeutic efficacy of a drug-like small molecule by predicting how it modulates a cancer pathway, can be incorporated for effective LSO of data-driven models for GMD.

Pathway-Guided Optimization of Deep Generative Molecular Design Models for Cancer Therapy

TL;DR

This work proposes how mechanistic models - such as pathway models described by differential equations - can be used for effective latent space optimization of JTVAEs and other similar models for GMD, and shows how a pharmacodynamic model can be incorporated for effective LSO of data-driven models for GMD.

Abstract

The data-driven drug design problem can be formulated as an optimization task of a potentially expensive black-box objective function over a huge high-dimensional and structured molecular space. The junction tree variational autoencoder (JTVAE) has been shown to be an efficient generative model that can be used for suggesting legitimate novel drug-like small molecules with improved properties. While the performance of the generative molecular design (GMD) scheme strongly depends on the initial training data, one can improve its sampling efficiency for suggesting better molecules with enhanced properties by optimizing the latent space. In this work, we propose how mechanistic models - such as pathway models described by differential equations - can be used for effective latent space optimization(LSO) of JTVAEs and other similar models for GMD. To demonstrate the potential of our proposed approach, we show how a pharmacodynamic model, assessing the therapeutic efficacy of a drug-like small molecule by predicting how it modulates a cancer pathway, can be incorporated for effective LSO of data-driven models for GMD.

Paper Structure

This paper contains 16 sections, 2 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Methods
  4. Unweighted JTVAE training
  5. Periodic weighted JTVAE training
  6. Data weighting
  7. Periodic retraining
  8. Therapeutic Score Calculation
  9. PARP1 Inhibition Model
  10. Therapeutic Score Calculation Using Rule-based Pathway-Guided Mechanistic Model
  11. Results & Discussion
  12. Limitations of Proposed Approach
  13. Limitations of Periodic Weighted Retraining of JTVAE
  14. Limitations of PARP1 Inhibition Model
  15. Limitations of Rule-based Pathway-guided Mechanistic Model
  16. ...and 1 more sections

Figures (6)

  • Figure 1: Overview of methodology
  • Figure 2: Relationship between therapeutic score and pIC50 according to the pathway guided mechanistic models.
  • Figure 3: Extended contact map of pathway model. There are 14 molecules arranged for their location in the nucleus (white background) and the cytoplasm (shaded background). Nineteen reaction rules are depicted by the lines with and without arrows which further define binding and/or catalytic reactions.
  • Figure 4: Distribution of therapeutic scores of generated molecules in consecutive retraining iterations for physiologically (a)viable, (b)modified, and (c)impractical pathway model.
  • Figure 5: Distribution of pIC50 of generated molecules in consecutive retraining iterations.
  • ...and 1 more figures