Artificial Intelligence creating new drugs from scratch by efficiently searching huge molecular possibilities

AI is revolutionizing medicine including radiology, pathology, and other medical specialties. Deep learning (DL) technologies are beginning to find applications in drug discovery including areas of molecular docking, transcriptomics, reaction mechanism elucidation, and molecular energy prediction.

A crucial step in many new drug discovery projects is the formulation of a well-motivated hypothesis for new lead compound generation (de novo design) or compound selection from available or synthetically feasible chemical libraries based on the available structure-activity relationship (SAR) data. The design hypotheses are often biased toward preferred chemistry or driven by model interpretation. Automated approaches for designing compounds with the desired properties de novo have become an active field of research in the last 15 years. The diversity of synthetically feasible chemicals that can be considered as potential drug-like molecules was estimated to be between 1030 and 1060. Great advances in computational algorithms, hardware, and high-throughput screening technologies notwithstanding, the size of this virtual library prohibits its exhaustive sampling and testing by systematic construction and evaluation of each individual compound. Local optimization approaches have been proposed, but they do not ensure the optimal solution, as the design process converges on a local or “practical” optimum by stochastic sampling or restricts the search to a defined section of chemical space that can be screened exhaustively.

Researchers have designed and implemented a novel computational strategy for de novo design of molecules with desired properties termed ReLeaSE (Reinforcement Learning for Structural Evolution). On the basis of deep and reinforcement learning (RL) approaches, ReLeaSE integrates two deep neural networks—generative and predictive—that are trained separately but are used jointly to generate novel targeted chemical libraries. ReLeaSE uses simple representation of molecules by their simplified molecular-input line-entry system (SMILES) strings only. Generative models are trained with a stack-augmented memory network to produce chemically feasible SMILES strings, and predictive models are derived to forecast the desired properties of the de novo–generated compounds. In the first phase of the method, generative and predictive models are trained separately with a supervised learning algorithm. In the second phase, both models are trained jointly with the RL approach to bias the generation of new chemical structures toward those with the desired physical and/or biological properties. In the proof-of-concept study, we have used the ReLeaSE method to design chemical libraries with a bias toward structural complexity or toward compounds with maximal, minimal, or specific range of physical properties, such as melting point or hydrophobicity, or toward compounds with inhibitory activity against Janus protein kinase 2. The approach proposed herein can find a general use for generating targeted chemical libraries of novel compounds optimized for either a single desired property or multiple properties.

13 thoughts on “Artificial Intelligence creating new drugs from scratch by efficiently searching huge molecular possibilities”

  1. I remember threads like this many years ago on the topic of computer vision. Smart and cranky people poopoo’ing neural networks in favor of decades-old feature engineering. And look where we are now. I imagine something similar will happen with computational drug discovery. And, as in computer vision, democratization of the technology will be the catalyst.

  2. I remember threads like this many years ago on the topic of computer vision. Smart and cranky people poopoo’ing neural networks in favor of decades-old feature engineering. And look where we are now. I imagine something similar will happen with computational drug discovery. And as in computer vision democratization of the technology will be the catalyst.

  3. Depends what you mean by work. Did the chemical produced react with the stick of protein poking out of the cell that was the target that the researchers thought would cause… blood clotting or whatever? Yes. I think some of the projects could achieve this. Did the chemical produced give you something that a human could ingest (ideally) or inject (if necessary) and it would a) React with said bit of protein b) Actually cause said blood clotting as planned c) Not be chewed up by the liver so you needed a new injection every 15 minutes. d) Not cause some other horrible thing to occur like shutting down the kidneys or poking holes in the brain. e) Not end up broken down by the liver into some toxin that caused more trouble than the original problem. f) Not show up as a blood clotter so that the body acted to reinforce the blood thinning mechanisms to counteract this alien attack on the blood system which ended up overcompensating g) Not have an effective therapeutic dose that was 85% of the overdose level, so that, given the variability between people, or even the same person of different days, you would be skating down a razors edge trying to medicate someone. h) You see where I am going here?

  4. Depends what you mean by work.Did the chemical produced react with the stick of protein poking out of the cell that was the target that the researchers thought would cause… blood clotting or whatever? Yes. I think some of the projects could achieve this.Did the chemical produced give you something that a human could ingest (ideally) or inject (if necessary) and it would a) React with said bit of proteinb) Actually cause said blood clotting as plannedc) Not be chewed up by the liver so you needed a new injection every 15 minutes.d) Not cause some other horrible thing to occur like shutting down the kidneys or poking holes in the brain.e) Not end up broken down by the liver into some toxin that caused more trouble than the original problem.f) Not show up as a blood clotter so that the body acted to reinforce the blood thinning mechanisms to counteract this alien attack on the blood system which ended up overcompensatingg) Not have an effective therapeutic dose that was 85{22800fc54956079738b58e74e4dcd846757aa319aad70fcf90c97a58f3119a12} of the overdose level so that given the variability between people or even the same person of different days you would be skating down a razors edge trying to medicate someone.h) You see where I am going here?

  5. Are you saying that while they looked like they would work in theory, in practice the compounds they produced did not work as predicted ?

  6. Are you saying that while they looked like they would work in theory in practice the compounds they produced did not work as predicted ?

  7. Over the last couple of decades, a number of similar sounding drug predicting by computer systems have been trialed. Sadly none of them have actually managed to move the needle (any needle) when it comes to how many drugs actually come out of the release-to-market end of the pipeline. Eventually they are bound to work. But it’s starting to a look a bit like fusion research.

  8. Over the last couple of decades a number of similar sounding drug predicting by computer systems have been trialed. Sadly none of them have actually managed to move the needle (any needle) when it comes to how many drugs actually come out of the release-to-market end of the pipeline.Eventually they are bound to work. But it’s starting to a look a bit like fusion research.

  9. I remember threads like this many years ago on the topic of computer vision. Smart and cranky people poopoo’ing neural networks in favor of decades-old feature engineering. And look where we are now.

    I imagine something similar will happen with computational drug discovery. And, as in computer vision, democratization of the technology will be the catalyst.

  10. Depends what you mean by work.

    Did the chemical produced react with the stick of protein poking out of the cell that was the target that the researchers thought would cause… blood clotting or whatever? Yes. I think some of the projects could achieve this.

    Did the chemical produced give you something that a human could ingest (ideally) or inject (if necessary) and it would
    a) React with said bit of protein
    b) Actually cause said blood clotting as planned
    c) Not be chewed up by the liver so you needed a new injection every 15 minutes.
    d) Not cause some other horrible thing to occur like shutting down the kidneys or poking holes in the brain.
    e) Not end up broken down by the liver into some toxin that caused more trouble than the original problem.
    f) Not show up as a blood clotter so that the body acted to reinforce the blood thinning mechanisms to counteract this alien attack on the blood system which ended up overcompensating
    g) Not have an effective therapeutic dose that was 85% of the overdose level, so that, given the variability between people, or even the same person of different days, you would be skating down a razors edge trying to medicate someone.
    h) You see where I am going here?

  11. Over the last couple of decades, a number of similar sounding drug predicting by computer systems have been trialed. Sadly none of them have actually managed to move the needle (any needle) when it comes to how many drugs actually come out of the release-to-market end of the pipeline.

    Eventually they are bound to work. But it’s starting to a look a bit like fusion research.

Comments are closed.