A new study has shown how machine learning can be used to simulate drug–target binding affinity to identify potential new drugs.
Researchers at Wuhan University (China) have developed a novel drug discovery platform called fingerprint-embedding framework for drug–target binding affinity prediction (FingerDTA), which uses ‘fingerprints’ of drugs and targets to predict how they would interact. Due to the costly, time-consuming nature of the in vivo drug discovery pipeline, developing a large-scale AI framework will be a powerful tool for virtual drug discovery.
Two virtual drug discovery methods are already in use: high-throughput screening and molecular-docking simulation. Both have proven successful methods; however, they require complex experimental design and evaluation. This means they aren’t suitable methods for screening drugs on a large scale. A third method focuses on modeling drug–target binding affinity. This method is more efficient and cost-effective than the other two, leading the current research team to build upon this work and improve its scalability.
FingerDTA combines a Convolutional Neural Network (CNN) model, to generate local information about the specific binding domain of a molecule, with the Simplified Molecular Input Line Entry System (SMILES) sequence, which generates ‘fingerprints’ of the drug molecule, while the full amino acid sequence of the target protein is used to generate a target fingerprint. This combination of local and global information allows FingerDTA to make predictions that more closely resemble biological interactions. Using these drug and target fingerprints as inputs, FingerDTA can then predict which potential drugs and targets will bind with the highest affinity.
AlphaFold computational modeling in drug discovery: the good, the bad and the AI
A new study has utilized AlphaFold machine-learning software to find potential targets in the Escherichia coli (E. coli) essential proteome for developing novel antibacterial drugs.
FingerDTA has lower data communication costs and utilizes approximate computing, meaning the system isn’t entirely dependent on memory space. Therefore, the system has serious potential in the drug discovery landscape, offering a virtual platform for large-scale drug–target binding affinity prediction that can save time, energy and resources while accelerating drug discovery research.
The team believes that this type of framework can be applied to find drugs that inhibit viruses from attaching to their targets. As senior author Juan Liu noted, “the FingerDTA can help discover some potential drugs for deactivating COVID-19 by binding to the spike target.”
In future, the researchers hope to use FingerDTA to analyze large amounts of data, eventually being able to develop real-world applications. “Our ultimate goal is to develop such technology and systems for users to tackle the application problems of analyzing extremely big data distributed in several data centers,” commented Liu.
The post Introducing FingerDTA: a new drug discovery framework appeared first on BioTechniques.
Full BioTechniques Article here
Powered by WPeMatico