Big pharma is turning to AI to save money, time, and effort on new drug R&D
The race continues to find the right drug to thoroughly mitigate the effects and spread of Covid-19. While some hope for herd immunity, the pharmaceutical industry is investing unfathomable resources in drug discovery.
Drug discovery is an expensive, complex, time-consuming process that can take up to 10 years, cost billions of dollars, and demand hundreds of people’s commitment before there is even the slightest chance of finding an effective treatment. The first step in this process is identifying which drugs are likely to treat a certain disease or condition. This requires extensive research, data analysis, and trial-and-error drug design efforts. It also relies on scientists’ expertise at predicting what might work based on their knowledge of biological processes from other diseases or similar compounds that have been tested for efficacy in similar conditions.
With recent advances in artificial intelligence (AI), it may soon become possible for computers to make these predictions faster than human scientists could ever hope to do so.
The application of drug discovery and AI
Transformer-based neural network architectures are changing the way we understand how data can be effectively processed. With self-supervised training methods, and access to massive data sets, researchers no longer need to manually label examples during pre-training and these models have been shown equally adept at learning both chemistry rules as well as grammar in languages.
Deep generative models, such as variational autoencoders and generative adversarial networks, are considered promising for the computational creation of novel molecules due to their state-of-the-art results in virtual synthesis of images, text, speech, and image captions. Virtual creation of new lead candidates requires exploration and performing a multiobjective optimization in a vast chemical space; the model needs to assess critical factors including drug activity (ease), selectivity (activity) toxicity...etc., all the while balancing between them using deep learning techniques that have been shown successful at synthesizing content like images or voice clips from scratch.
Let's say you're designing a drug candidate for an unknown target in the SARS-CoV. You don't have access to labeled data and your engineering team is too busy. In this case, it might be worth looking into Bayesian optimization or conditional generative models that can help provide useful insights without sacrificing efficiency. AI systems like SEDGE employ a variety of techniques, including machine learning and deep neural networks, which can learn complex patterns in data. SEDGE is capable of predicting what might work based on data from biological processes from other diseases or similar compounds that have been tested.
Using AI to make complex drug-discovery processes simpler
The use of AI in streamlining the complex processes required in finding a drug for Covid-19 is proving beneficial. The availability of large databases allows researchers to pretrain AI models that understand chemical structure, foregoing the need for hand-labeled data.
AI tools help researchers perform attribute-controlled generation of antimicrobial peptides (AMPs) by using the latent features from a pre-trained autoencoder. Using deep learning (DL) and high-throughput molecular simulations when performing additional in silico screening - the process of sifting through more of the vast chemical space and optimizing pharmacological properties before moving on to expensive and time-consuming lab testing - allows for the creation of a number of designed novel peptides for wet lab synthesis and validation. From this it is easier and faster for researchers to confirm potency and toxicity.
Equipped with a statistical understanding of chemistry, the AI tool can be positioned towards a number of downstream tasks, including predicting how chemicals will react with each other, and outputting new molecular structures. This output can provide innovative ideas for drug researchers by suggesting molecules that do not currently exist in a database, but could potentially be drug candidates.
Additionally, vast amounts of patient records can be used to train models in quickly identifying ideal patients for lifesaving clinical trials. This can be used to further understanding and measure the effects and side-effects of a new drug, vaccine or treatment.
SEDGE hails in a new era of accelerated drug discovery
At SEDGE, we have built robust and scalable AI tools and platforms that support multi-disciplinary arenas of discovery. This saves time, money and effort.
Using the power of AI and machine learning, SEDGE can accelerate how pharmaceutical companies perform complex scientific discovery tasks. This would empower the rapid design and optimization of novel materials and molecules, and successfully move these organizations into this new phase of accelerated drug discovery.
If you're looking for an innovative company to help speed up your next scientific project, don't hesitate to reach out. You’re able to try SEDGE before buying so there’s no risk involved. We look forward to hearing from you soon and helping you accelerate your drug discovery for Covid-19 through AI.
Comments