Existing drug discovery pipelines take 5–10 years and cost billions of dollars. Computational approaches aim to sample from regions of the whole molecular and solid-state compounds called chemical space, which could be on the order of$10^{60}$. Deep generative models can model the underlying probability distribution of both the physical structures and property of drugs and relate them nonlinearly. By exploiting patterns in massive datasets, these models can distill salient features that characterize the molecules. Generative adversarial networks (GANs) discover drug candidates by generating molecular structures that obey chemical and physical properties and show affinity toward binding with the receptor for a target disease. However, classical GANs cannot explore certain regions of the chemical space and suffer from training instabilities. The practical utility of such models is limited due to the vast size of the search space, characterized by millions of parameters. A full quantum GAN may require more than 90 qubits even to generate small molecules with up to nine heavy atoms. The proposed quantum GAN with a hybrid generator (QGAN-HG) model is composed of a hybrid quantum generator that supports various number of qubits and quantum circuit layers, and a classical discriminator. The QGAN-HG with less than 20% of the original parameters can learn molecular distributions as efficiently as its classical counterpart. Another extended version of the proposed QGAN-HG, which utilizes multiple quantum subcircuits, considerably accelerates our standard QGAN-HG training process and avoids the potential gradient vanishing issue of deep neural networks.
Existing drug discovery pipelines take 5–10 years and cost billions of dollars. Computational approaches aim to sample from regions of the whole molecular and solid-state compounds called chemical space, which could be on the order of1060. Deep generative models can model the underlying probability distribution of both the physical structures and property of drugs and relate them nonlinearly. By exploiting patterns in massive datasets, these models can distill salient features that characterize the molecules. Generative adversarial networks (GANs) discover drug candidates by generating molecular structures that obey chemical and physical properties and show affinity toward binding with the receptor for a target disease. However, classical GANs cannot explore certain regions of the chemical space and suffer from training instabilities. The practical utility of such models is limited due to the vast size of the search space, characterized by millions of parameters. A full quantum GAN may require more than 90 qubits even to generate small molecules with up to nine heavy atoms. The proposed quantum GAN with a hybrid generator (QGAN-HG) model is composed of a hybrid quantum generator that supports various number of qubits and quantum circuit layers, and a classical discriminator. The QGAN-HG with less than 20% of the original parameters can learn molecular distributions as efficiently as its classical counterpart. Another extended version of the proposed QGAN-HG, which utilizes multiple quantum subcircuits, considerably accelerates our standard QGAN-HG training process and avoids the potential gradient vanishing issue of deep neural networks.
Current drug discovery processes can cost billions of dollars and usually take five to ten years. People have been researching and implementing various computational approaches to search for molecules and compounds from the chemical space, which can be on the order of 1060 molecules. One solution involves deep generative models, which are artificial intelligence models that learn from nonlinear data by modeling the probability distribution of chemical structures and creating similar data points from the trends it identifies. These generative models can extract salient features that characterize the molecules. However, they often suffer from increased time complexity. Aiming for faster runtime and greater robustness when analyzing high-dimensional data, we designed and implemented a Hybrid Quantum-Classical Generative Adversarial Network (QGAN) to synthesize molecules. There are two parts in the QGAN: a quantum generator that creates molecules based on the probability distributions of likely combinations and a classical discriminator that differentiates real molecules from the generated ones. We hypothesized that a quantum generator would be more impactful, offering a runtime speedup and increasing generated molecule possibilities, because we could use quantum mechanical phenomena superposition, entanglement, and interference to analyze more atom-bond combinations than a classical generator. The PyTorch-PennyLane implementation of the QGAN generated seven chemically stable molecules out of 300, a 2.3% success rate. Although the QGAN is a work in progress, it has demonstrated a path towards more efficient drug development, which would accelerate the development of medicines and reduce costs for the whole R&D process.
Traditional drug discovery pipeline takes several years and cost billions of dollars. Deep generative and predictive models are widely adopted to assist in drug development. Classical machines cannot efficiently produce atypical patterns of quantum computers which might improve the training quality of learning tasks. We propose a suite of quantum machine learning techniques e.g., generative adversarial network (GAN), convolutional neural network (CNN) and variational auto-encoder (VAE) to generate small drug molecules, classify binding pockets in proteins, and generate large drug molecules, respectively.
Contemporary drug discovery and development processes require billions of dollars and lengthy amounts of time, which is why researchers are utilizing computational chemistry methods like machine learning to speed up molecular synthesis pathways. Generative Adversarial Networks, in particular, have shown immense potential for accelerating the drug discovery pipeline. However, these models consume a significant amount of computational resources, as they must perform adversarial training on expansive datasets to accurately sample the vast chemical space. Aiming for more efficient runtimes and rigorous analyses of high-dimensional molecular data with quantum computing, we previously developed a hybrid quantum-classical generative adversarial network (QGAN) to generate potential small organic drug candidates. Later, inspired by the classical NetGAN, we upgraded QGAN to a hybrid quantum-classical graph generative adversarial network (QNetGAN) which generated graph structures via random walks. This model was more efficient and had a higher accuracy rate (47.0%) than QGAN (2.3%). Most recently, we developed QNetGAN v2, an upgraded version of QNetGAN that features graph convolution. This model was able to generate 273 out of 300 structurally valid molecules that satisfy the octet rule and Lipinski's Rule of Five, yielding an 91.0% success rate and a training time of 35 minutes. This was made possible by integrating several post-processing algorithms into the original QNetGAN model, including the octet rule satisfaction validator, hydrogen addition algorithm, formal charge calculation, and molecular geometry (.XYZ file coordinate) calculation. In the future, we plan to use the QNetGAN v2 pathway to generate molecules with a wider range of elements and add functionality to calculate important molecular properties, ideally developing a third and more capable version of the QNetGAN molecular generation pathway that is able to outperform the accuracy of classical NetGAN.
Traditional drug discovery pipelines can require multiple years and billions of dollars of investment. Deep generative and discriminative models are widely adopted to assist in drug development. Classical machines cannot efficiently reproduce the atypical patterns of quantum computers, which may improve the quality of learned tasks. We propose a suite of quantum machine learning techniques: incorporating generative adversarial networks (GAN), convolutional neural networks (CNN) and variational auto-encoders (VAE) to generate small drug molecules, classify binding pockets in proteins, and generate large drug molecules, respectively.
Drug discovery is a notoriously lengthy, costly, and often unsuccessful process. Traditional methods rely heavily on high-throughput screening and iterative optimization, which are computationally intensive and time-consuming. This research investigates the potential of Quantum Generative Adversarial Networks (QGANs) to significantly accelerate this process by learning efficient molecular representations and generating novel drug candidates. QGANs utilize quantum superposition to explore a vast chemical space, offering a fundamentally different approach compared to classical generative models. We propose a framework where a quantum generator network learns to create molecular structures, while a classical discriminator network evaluates their properties against desired criteria. The core innovation lies in the application of quantum computing principles to molecular representation learning, aiming to drastically reduce the computational burden associated with generating and evaluating potential drug candidates. The resulting molecular representations are then used to guide the generation of new compounds with optimized characteristics, addressing a critical bottleneck in the drug discovery pipeline. The framework presented herein provides a theoretical foundation and outlines a potential approach for leveraging quantum computing to revolutionize drug discovery.
Drug discovery for a given protein target is a complex, resource-intensive process, often requiring years of research and billions of dollars in investment. Computational techniques, particularly deep learning models, have accelerated molecular generation, but they remain constrained by classical computational limitations and reliance on known chemical patterns. To address these challenges, we propose a hybrid quantum conditional GAN with gradient penalty for drug design, integrating quantum computing to enhance exploration of the vast chemical space. The proposed approach leverages a hybrid classical and quantum generative model to create novel molecular structures while ensuring they adhere to chemical and pharmacological properties. By conditioning the generative model on protein target sequences, we improve molecular specificity, increasing the likelihood of discovering viable drug candidates. Our framework generates chemically valid, diverse, and target-specific molecules, offering a promising avenue for more efficient and precise data-driven drug discovery.