Abstract 3380: Predicting the effects of small molecules on transcriptome of cancer cell lines using deep learning

Developing a computational model that can predict the effects of small molecules on biological status of a cell can be very useful for various applications, including candidate drug compound screening and toxicity testing in silico. In this study, we developed a deep learning model that predicts the effects of small molecules on gene expression levels of a cell. This deep learning model takes two inputs: the structure of a chemical, presented in SMILES, and a protein sequence encoded by a gene of interest. The model processes these two input data using a fingerprinting method for the input chemical structure and a convolutional neural network for the input protein sequence. As a result, the model predicts whether a given molecule upregulates or downregulates a gene of interest where the molecule and gene are both provided as inputs. The deep learning model was developed by using so called L1000 profiles covering gene expression levels of various cancer cell lines under a large number of perturbation conditions [1]. L1000 is a cost-effective transcriptome technology that can accurately infer the expression levels of 9,196 genes on the basis of expression levels of 978 ‘landmark9 genes that are directly measured. In this study, the deep learning model was developed for three different cancer cell lines, HA1E (kidney), HCC515 (lung), and HEPG2 (liver), by using L1000 profiles covering expression levels of the landmark genes individually perturbed with 5,213 molecules for the HA1E cell line, 5,171 molecules for the HCC515, and 3,529 for HEPG2. The deep learning models developed in this study will be useful for a wide range of studies that examine the effects of small molecules, for example in drug development and toxicity testing. [1] Subramanian et al. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles. Cell 171, 1437-1452 (2017) Citation Format: Junhyeok Jeon, Sang Mi Lee, GaRyoung Lee, Hyun Uk Kim. Predicting the effects of small molecules on transcriptome of cancer cell lines using deep learning [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3380.