PharmaNet: Pharmaceutical discovery with deep recurrent neural networks

Abstract

The discovery and development of novel pharmaceuticals is an area of active research mainly due to the large investments required and long payback times. As of 2016, the development of a novel drug candidate required up to $ USD 2.6 billion in investment for only 10% rate of approval by the FDA. To help decreasing the costs associated with the process, a number of in silico approaches have been developed with relatively low success due to limited predicting performance. Here, we introduced a machine learning-based algorithm as an alternative for a more accurate search of new pharmacological candidates, which takes advantage of Recurrent Neural Networks (RNN) for active molecule prediction within large databases. Our approach, termed PharmaNet was implemented here to search for ligands against specific cell receptors within 102 targets of the DUD-E database, which contains 22886 active molecules. PharmaNet comprises three main phases. First, a SMILES representation of the molecule is converted into a raw molecular image. Second, a convolutional encoder processes the data to obtain a fingerprint molecular image that is finally analyzed by a Recurrent Neural Network (RNN). This approach enables precise predictions of the molecules’ target on the basis of the feature extraction, the sequence analysis and the relevant information filtered out throughout the process. Molecule Target prediction is a highly unbalanced detection problem and therefore, we propose that an adequate evaluation metric of performance is the area under the Normalized Average Precision (NAP) curve. PharmaNet largely surpasses the previous state-of-the-art method with 97.7% in the Receiver Operating Characteristic curve (ROC-AUC) and 65.5% in the NAP curve. We obtained a perfect performance for human farnesyl pyrophosphate synthase (FPPS), which is a potential target for antimicrobial and anticancer treatments. We decided to test PharmaNet for activity prediction against FPPS by searching in the CHEMBL data set. We obtained three (3) potential inhibitors that were further validated through both molecular docking and in silico toxicity prediction. Most importantly, one of this candidates, CHEMBL2007613, was predicted as a potential antiviral due to its involvement on the PCDH17 pathway, which has been reported to be related to viral infections.

Molecular Toxicity Virtual Screening Applying a Quantized Computational SNN-Based Framework

“…This advancement may offer an opportunity for integrated hardware and software solutions to solve complex in silico tasks of medical chemistry. Indeed, one approach followed by some authors [ 63 ] was to use recurrent neural networks on bi-dimensional fingerprint-like representations of atoms and bonds. Consequently, technological progress in neuromorphic computing could provide further applications to chemoinformatics, employing systems that reflect the mechanisms of brain activity and thus more explainable than standard “black-box” approaches derived from ANNs.…”

Institutional change and stability in strategic networks in the Brazilian pharmaceutical industry

“…However, different from other studies that consider the actors as a driver of institutional change (Kingston and Caballero, 2009), this study highlights the influence of the exogenous and endogenous shocks in the institutional design in each track (Figure 2) because institutions are shapers of organizational arrangements (Jepperson, 1991). If the context-dependent nature of institutions (Fortwengel and Jackson, 2016) works as a representation of a moving equilibrium path, institutional change is a new moving equilibrium path (Aoki, 2001) to deal with challenges in the pharmaceutical industry such as sustainable supply chains, performance measurement, lean and agile operation, cold chain management and risk management (Ding, 2018; Dixit et al , 2019; Ruiz Puentes et al , 2021).…”

“…Moreover, the function and change of the institutions are relevant elements to the process logic of consensus building in SNs (Blatter, 2003) because the distinction between formal and informal institutions is key to understanding institutional change (Paz, 2015). Consequently, pharmaceutical firms play an important role in setting market and timing rules for their products because they affect peoples’ lives (Ruiz Puentes et al , 2021).…”

Predicting target–ligand interactions with graph convolutional networks for interpretable pharmaceutical discovery

“…This working objective stresses the importance of appropriate molecular structure representation. In this respect, previous approaches have employed two-dimensional molecular images built from one-hot embeddings over atoms, a general method to vectorize categorical features, and have been analyzed under convolutional neural networks (CNN) 18 and recurrent neural networks (RNN) 19 to predict TLIs. However, these contributions disregard molecular structure, as well as the target protein's information.…”

“…Our contributions are three-fold: (1) we model the spatial configuration of both the target protein and the ligand through bidirected graphs, (2) we integrate relevant chemical/primary structure information from proteins for TLI prediction, and (3) we propose a gradient-based method to compute adversarial molecule augmentations that preserve relevant biological backgrounds and improve both interpretability and overall performance. We train PLA-Net models for 102 pharmacologically-relevant protein targets and establish the new state-of-the-art, outperforming the previous one 19 by a large margin of 19.8% on mean average precision (mAP) in a curated version of the actives as decoys (AD) dataset. Moreover, we perform a virtual screening between molecules of two large datasets (ChEMBL 34 , and Drug Repurposing Hub 35 ) with the perfect-scoring targets, and corroborate that our method accurately predicts experimentally validated TLIs.…”