PharmacoDB: an integrative database for mining in vitro anticancer drug screening studies

Abstract

Recent cancer pharmacogenomic studies profiled large panels of cell lines against hundreds of approved drugs and experimental chemical compounds. The overarching goal of these screens is to measure sensitivity of cell lines to chemical perturbations, correlate these measures to genomic features, and thereby develop novel predictors of drug response. However, leveraging these valuable data is challenging due to the lack of standards for annotating cell lines and chemical compounds, and quantifying drug response. Moreover, it has been recently shown that the complexity and complementarity of the experimental protocols used in the field result in high levels of technical and biological variation in the in vitro pharmacological profiles. There is therefore a need for new tools to facilitate rigorous comparison and integrative analysis of large-scale drug screening datasets. To address this issue, we have developed PharmacoDB (pharmacodb.pmgenomics.ca), a database integrating the largest cancer pharmacogenomic studies published to date. Here, we describe how the curation of cell line and chemical compound identifiers maximizes the overlap between datasets and how users can leverage such data to compare and extract robust drug phenotypes. PharmacoDB provides a unique resource to mine a compendium of curated cancer pharmacogenomic datasets that are otherwise disparate and difficult to integrate.

BDKANN - Biological Domain Knowledge-based Artificial Neural Network for drug response prediction

“…2) The Cancer Therapeutic Response Portal (CTRPv2) dataset [18], consisting of more than 800 pan-cancer cell lines, screened with 481 targeted and chemotherapy drugs. We employed GDSC dataset for training and CTRPv2 dataset for test and obtained them via Pharma-coGx R package [19] and PharmacoDB [10]. These two datasets have 93 drugs in common, however, we focused on 6 of those drugs including Doxorubicin, Tamoxifen, Masitinib, 17-AAG, GDC-0941, and PLX4720 because these drugs had pathway information available in REACTOME and we filtered out drugs with a high fraction of NA values for the outcome.…”

“…These pre-clinical datasets are larger than clinical datasets and are screened with tens or hundreds of drugs which makes them useful resources for training a computational model. For drug response prediction on pre-clinical datasets, previous studies have proposed different methods such as regression [7], kernel learning [8], and deep neural networks (DNNs) [9] to map the genomic data-often gene expression-to a measure of response such as the IC50 or area above the dose-response curve (AAC) [10]. Although these methods have shown promising results, they have not considered prior biological knowledge in their models.…”

AITL: Adversarial Inductive Transfer Learning with input and output space adaptation for pharmacogenomics

“…To study AITL's performance, similar to (Geeleher et al, 2017;Sharifi-Noghabi et al, 2019b), we employ the model trained on Docetaxel, Paclitaxel, or Bortezomib to predict the response for patients in several TCGA cohorts for which no drug response was recorded. For each drug, we extract the list of annotated target genes from the PharmacoDB resource (Smirnov et al, 2017). We excluded Cisplatin because there was only one annotated target gene for it in PharmacoDB.…”

Karyotypic divergence confounds cellular phenotypes in large pharmacogenomic studies

“…We analyzed 3,631 Affymetrix Genome-Wide Human SNP 6.0 (Affy6) arrays and 668 HumanOmni2.5-Quad (Omni) arrays from 1,691 cell lines profiled by a compendium of 6 pharmacogenomic studies (Table 1). We downloaded files using accession codes outlined in Table 1 and obtained cell line annotations from PharmacoDB (version 1.0.0) 23,24 . We removed samples if they failed quality control metrics specific for Affy6 or Omni platforms 25 .…”