Methods and systems for testing molecules and their properties - Patents.com

Abstract

Provided herein is a method for testing a molecule and its properties, the method comprising: a) obtaining a request including a representation of at least one property of a molecule and a corresponding task; b) performing inference using the representation on at least one machine learning (ML) model, the at least one ML model configured to generate a final result of the inference by mimicking a result of the task; c) performing an inference confidence test on the final result of the inference, wherein i) in response to a good inference confidence test, the final result of the inference is used to obtain a task result, and ii) in response to a bad inference confidence test, the task is performed using the representation to obtain a task result; and d) outputting the task result.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application No. 63 / 314,971, filed February 28, 2022, the disclosure of which is incorporated herein by reference in its entirety. [Background technology]

[0002] Computational chemistry has become an established tool for the molecular and materials discovery process in many areas of industry. It can provide highly accurate prediction of chemical phenomena and exploration of molecular properties that are not accessible from experiments alone and / or require a great deal of effort. Several computational chemistry platforms exist to facilitate highly accurate prediction of chemical behavior of molecules and materials. Some examples of such platforms include Gaussian™ (https: / / gaussian.com), Q-Chem™ (https: / / www.q-chem.com), and Schrodinger™ (https: / / www.schrodinger.com). For these platforms, improving computational efficiency is one of the most essential development activities that helps to accelerate the materials discovery process, and platform developers have made great efforts to improve efficiency without sacrificing the accuracy level of computational predictions. Examples are described in the efficiency improvement section at https: / / gaussian.com / gdiffs / .

[0003] However, since the efficiency of these platforms is improved based on their development roadmap, users must tolerate the efficiency level provided by the computational approach that meets their accuracy needs until a new package with improved efficiency is released, and / or users must choose a less accurate but more efficient computational approach for their efficiency needs. This causes undesired delays in the progress of users' research and / or failure to make high-precision predictions due to the use of less accurate approaches. There are several open source packages for computational chemistry, such as PySCF (https: / / pyscf.org) and Psi4 (https: / / psicode.org), where users can directly contribute to the development and therefore the efficiency level of the computational chemistry methods in the packages. However, improving the efficiency of the methods requires a lot of effort and deep knowledge and experience in computational chemistry methods. Summary of the Invention

[0004] The present disclosure recognizes the need for an improved method and system that overcomes at least one of the shortcomings set forth above.The present disclosure provides a method and system for testing molecules and their properties.The present disclosure may improve existing methods and systems in at least some aspects.

[0005] One or more embodiments of the methods and systems disclosed herein may advantageously provide a computational chemistry platform that provides improved computational efficiency by allowing users to run computational chemistry calculations / pipelines on the platform.

[0006] One or more embodiments of the methods and systems disclosed herein may advantageously provide a computational chemistry platform that improves efficiency more frequently as a user runs computational chemistry calculations / pipelines on the platform more frequently.

[0007] One or more embodiments of the methods and systems disclosed herein may advantageously provide a computational chemistry platform that becomes more efficient as a user runs more computational chemistry calculations / pipelines on the platform.

[0008] One or more embodiments of the methods and systems disclosed herein may advantageously provide a computational chemistry platform that can improve the efficiency of computational chemistry methods without the need for human labor.

[0009] One or more embodiments of the methods and systems disclosed herein may have the advantage of providing a computational chemistry platform that can improve the efficiency of computational chemistry methods without extensive knowledge or experience in improving their efficiency.

[0010] One or more embodiments of the methods and systems disclosed herein may advantageously provide a computational chemistry platform that can improve the efficiency of conducting laboratory experiments.

[0011] One or more embodiments of the methods and systems disclosed herein may have another advantage in that they provide a framework that may be realized by combining machine learning techniques with a variety of computational chemistry techniques, computational discovery pipelines, and chemical discovery toolkits that may also consist of laboratory experiments.

[0012] In an aspect, the disclosure provides a method for testing a molecule and a property thereof, the method comprising: a) obtaining a request including a representation of at least one property of the molecule and a corresponding task; b) performing inference using the representation for at least one machine learning (ML) model configured to generate an inference final outcome by mimicking a result of the task; c) performing an inference confidence test on the final outcome of the inference, i) obtaining a task result using the final outcome of the inference corresponding to a good inference confidence test; and ii) obtaining a task result by performing the task using the representation corresponding to a bad inference confidence test; and d) outputting the task result.

[0013] Another aspect of the disclosure provides a non-transitory computer-readable medium that includes machine-executable code that, upon execution by one or more computer processors, performs any of the methods described above or elsewhere herein.

[0014] Another aspect of the present disclosure provides a system including one or more computer processors and a computer memory coupled thereto, the computer memory including machine-executable code that, upon execution by the one or more computer processors, performs any of the methods described above or elsewhere herein.

[0015] Additional aspects and advantages of the present disclosure will become readily apparent to those skilled in the art from the following detailed description. In the following detailed description, only exemplary embodiments of the present disclosure are shown and described. As will be recognized, the present disclosure is capable of other various embodiments, and its several details are capable of modification in various obvious respects, all without departing from the present disclosure. Thus, the drawings and description should be regarded as illustrative in nature, and not restrictive.

[0016] Incorporation by Reference All publications, patents, and patent applications mentioned herein are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. To the extent that a publication, patent, or patent application incorporated by reference conflicts with the disclosure contained herein, the present specification is intended to supersede and / or take precedence over any such conflicting material. [Brief description of the drawings]

[0017] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and in which the accompanying drawings (also referred to herein as "Figure" and "FIG.") are shown: [Figure 1] FIG. 1 is a diagram of a system for testing molecules and their properties according to some embodiments disclosed herein. [Diagram 2] FIG. 1 is a flow diagram of a method for testing molecules and their properties according to some embodiments disclosed herein. [Diagram 3] FIG. 1 is a flow diagram of an embodiment of a method for calculating ground state energies with DFT wB97X / 6-31g* level of accuracy based on the three-dimensional coordinates of a molecule. [Figure 4] FIG. 1 is a flow diagram of an embodiment of a method for computing ground-state optimized geometries with CCSD(T) / CBS level accuracy based on SMILES. [Diagram 5] FIG. 1 is a flow diagram of an embodiment of a method for obtaining barrier heights for a chemical reaction of interest using DFT B3LYP / cc-pvd based on the 3D coordinates of reactants and products. [Figure 6] FIG. 1 is a flow diagram of an embodiment of a method for calculating the UV-Vis spectrum of a molecule of interest using TDDFT PBE0 / 6-311+G(d,p) based on the three-dimensional coordinates of the molecule. [Figure 7] FIG. 1 is a flow diagram of an embodiment of a method for obtaining a list of organic molecules that have strong peaks at 350 nm and 500 nm in the UV-Vis spectrum and have less than 15 heavy atoms. [Figure 8] FIG. 1 is a flow diagram of an embodiment of a method for predicting the residence time of a ligand in a ligand-binding site of a protein of interest based on the three-dimensional coordinates of the binding conformation between the ligand and the protein. [Figure 9] FIG. 1 is a flow diagram of an embodiment of a method for identifying a ligand from a ligand database that has the highest binding free energy with a protein of interest PDB ID based on SMILES from a ligand library. [Figure 10] FIG. 1 is a flow diagram of an embodiment of a method for identifying self-complementary Janus motifs based on guanine and cytosine base pairs that form nanotubes with an outer diameter of 3.7 nm. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] While various embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It is understood that various alternatives to the embodiments of the present invention described herein may be used.

[0019] Whenever the terms "at least," "greater than," or "greater than or equal to" precede the first number in a series of two or more numbers, the terms "at least," "greater than," or "greater than or equal to" apply to every number in the series. For example, 1, 2, or 3 or more is equivalent to 1 or more, 2 or more, or 3 or more.

[0020] Whenever the terms "no more than," "less than," or "less than or equal to" precede the first number in a series of two or more numbers, the terms "no more than," "less than," or "less than or equal to" apply to each number in the series. For example, 3, 2, or 1 or less is equivalent to 3 or less, 2 or less, or 1 or less.

[0021] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs. As used in this specification and the appended claims, the singular forms "a", "an" and "the" include plural references unless the context clearly indicates otherwise. Any reference to "or" in this specification is intended to include "and / or" unless otherwise indicated.

[0022] The word "plurality" means "two or more" unless expressly specified otherwise.

[0023] The word "herein" means "in this application, including any matter which may be incorporated by reference," unless expressly specified otherwise.

[0024] The word "eg" and similar words mean "for example," and thus do not limit the described word or phrase. For example, in the sentence "A computer transmits data (eg (for example) instructions, data structures) over the Internet," the word "eg" describes that "instructions" are an example of "data" that a computer may transmit over the Internet, and also describes that "data structures" are an example of "data" that a computer may transmit over the Internet. However, both "instructions" and "data structures" are merely examples of "data," and other things besides "instructions" and "data structures" may be "data."

[0025] When values ​​are described as ranges, the disclosure includes disclosure of all possible subranges within such ranges, as well as specific numerical values ​​within such ranges, whether or not a specific numerical value or subrange is expressly recited.

[0026] In the following detailed description, reference is made to the accompanying drawings, which form a part of this specification. In the drawings, similar reference numerals typically designate similar components unless the context indicates otherwise. The exemplary embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments may be used and other changes may be made without departing from the scope of the subject matter presented herein. It will be readily understood that the aspects of the present disclosure as generally described and illustrated in the drawings herein can be arranged, substituted, combined, separated, and designed in a wide variety of configurations, all of which are expressly contemplated herein.

[0027] As used herein, the term "classical" in the context of computing or calculations generally refers to calculations performed with binary values ​​using discrete bits without the use of quantum mechanical superposition and entanglement. A classical computer may be a digital computer, such as a computer that uses discrete bits (e.g., 0s and 1s) without the use of quantum mechanical superposition and entanglement.

[0028] As used herein, the term "non-classical" used in the context of computing or computation generally refers to any method or system for performing computational procedures outside the classical computing paradigm.

[0029] As used herein, the term "quantum device" generally refers to any device or system for performing computations using any of the quantum mechanical phenomena, such as quantum mechanical superposition and quantum mechanical entanglement.

[0030] As used herein, the terms “quantum computation,” “quantum procedure,” “quantum operation,” and “quantum computer” generally refer to any method or system for performing computations using quantum mechanical operations (such as unitary transformations on quantum channels or completely positive definite trace-preserving (CPTP) maps) on Hilbert spaces represented by quantum devices.

[0031] As used herein, the term "noisy intermediate-scale quantum device" (NISQ) generally refers to any quantum device capable of performing tasks that exceed the performance of today's classical digital computers.

[0032] As used herein, the term "quantum chemistry calculation" generally refers to calculations for predicting electronic structures and molecular properties using quantum mechanics.

[0033] As used herein, "molecular mechanical calculation" generally refers to molecular modeling calculations based on classical mechanics.

[0034] As used herein, "computational chemistry calculation" generally refers to a computer simulation to aid in solving a chemical problem. Computational chemistry calculation may include quantum chemistry calculation. Computational chemistry calculation may further include molecular mechanics calculation.

[0035] As used herein, a "chemistry discovery toolkit" generally refers to a toolkit to aid in the design and discovery of new molecules or molecular properties. The chemistry discovery toolkit may include computational chemistry methods, molecular mechanics methods, and quantum chemistry methods. The chemistry discovery toolkit may further include a set of the above methods performed in a particular order (a pipeline). The chemistry discovery toolkit may further include a laboratory experimentation approach.

[0036] As used herein, the term "machine learning (ML) model" generally refers to a flexible algorithm that is trained to recognize certain patterns. ML algorithms include, but are not limited to, supervised learning, semi-supervised learning, unsupervised learning, transfer learning, delta learning, and reinforcement learning.

[0037] As used herein, the term "molecule" generally refers to a substance made up of one or more atoms. This can include condensed phase systems such as solids and liquids. It can further include materials.

[0038] As used herein, the term "molecular properties" generally refers to properties including chemical, physical, and structural properties of a molecule. It may also include pharmacological or biological properties of a molecule.

[0039] As used herein, the term "fingerprint" generally refers to a method for uniquely encoding the structure of a molecule.

[0040] As used herein, the term "ground state" generally refers to the lowest energy state of a system, which may include electronic, vibrational, and rotational energy states.

[0041] As used herein, the term "excited state" generally refers to a state having higher energy than the ground state, which can include electronic, vibrational, and rotational energy states.

[0042] As used herein, the term "molecular orbital" generally refers to a mathematical function that describes the positions and wave-like behavior of electrons in a molecule.

[0043] As used herein, the term "ionization potential" generally refers to the amount of energy required to remove one electron from a neutral atom or molecule.

[0044] As used herein, the term "electron affinity" generally refers to the amount of energy released when an electron is added to a neutral atom or molecule.

[0045] As used herein, the term "singlet-triplet gap" generally refers to the energy difference between the singlet and triplet spin states of a system.

[0046] As used herein, the term "atomic charge" generally refers to a real number that describes the distribution of electron density in a molecule.

[0047] As used herein, the term "dipole moment" generally refers to a measure of the separation of positive and negative charges within a system.

[0048] As used herein, the term "charge density" generally refers to the difference per unit volume between the positive charge of the atomic nuclei in a system and the distribution of negative charges carried by the electrons in the system.

[0049] As used herein, the term "spectroscopic properties" generally refers to properties characterized by spectroscopy, such as X-ray, infrared (IR), Raman, nuclear magnetic resonance (NMR), and electronic absorption / emission spectra.

[0050] As used herein, the term "binding affinity" generally refers to the strength of the binding interaction between a single molecule (e.g., a protein or DNA) and its ligand / binding partner (e.g., a drug or inhibitor).

[0051] As used herein, the term "equilibrium geometry" generally refers to the geometry of a molecule that corresponds to the true minimum on its respective potential energy surface.

[0052] As used herein, the term "reactivity" generally refers to the impetus of a chemical to undergo a chemical reaction.

[0053] As used herein, the term "hydrophobicity" generally refers to the physical property of molecules being seemingly repelled from a body of water.

[0054] As used herein, the term "synthesizability" generally refers to the ability to be synthesized by a particular synthetic route.

[0055] As used herein, the term "conformational entropy" generally refers to the entropy associated with the number of conformations of a molecule.

[0056] As used herein, the term "residence time" generally refers to the amount of time a molecule is bound to its target molecule.

[0057] As used herein, the term "protein structure" generally refers to the three-dimensional arrangement of atoms in an amino acid chain molecule.

[0058] As used herein, the term "protein pocket" generally refers to a cavity on a protein surface or within a protein that has suitable properties for the binding of a ligand.

[0059] As used herein, the term "protein-ligand interactions" generally refers to the behavior of interactions between proteins and ligands.

[0060] As used herein, the term "binding free energy" generally refers to the free energy change associated with the binding event of a molecule with its target molecule.

[0061] As used herein, the term "toxicity" generally refers to the degree to which a chemical or a particular mixture of substances can damage an organism.

[0062] As used herein, the term "catalyst molecules" generally refers to substances that help increase the rate of chemical reactions.

[0063] As used herein, the term "conformational search" generally refers to the process of finding energetically favorable conformations of a molecule.

[0064] As used herein, the term "barrier height" generally refers to the energy difference between the minimum and maximum values ​​on a potential energy diagram.

[0065] As used herein, the term "substrate" generally refers to a chemical species being observed in a chemical reaction.

[0066] As used herein, the term "ab initio molecular dynamics" generally refers to a methodology in which finite temperature dynamical trajectories are generated by using forces calculated on the fly from electronic structure calculations.

[0067] As used herein, the term "classical force field" generally refers to a set of functional forms and parameters used in the calculation of the potential energy of a system of atoms or molecules or coarse particles.

[0068] As used herein, the term "classical molecular dynamics" generally refers to a methodology in which finite temperature dynamical trajectories are generated by using forces calculated on-the-fly from classical force fields.

[0069] The present application discloses methods and systems for testing molecules and their properties.

[0070] Neither the title nor the abstract should be considered in any way limiting as to the scope of the disclosed invention. The title of this application and the section headings provided in this application are for convenience only and are not to be considered in any way limiting to the present disclosure.

[0071] Machine Learning Machine learning (ML) can be defined as training (e.g., adjusting parameters in) a flexible computer algorithm with a particular set of data. More specifically, ML includes supervised learning, semi-supervised learning, unsupervised learning, transfer learning, delta learning, and reinforcement learning. In supervised learning, parameters in an ML model are updated so that the model's output and the truly labeled data produce similar results. In unsupervised learning, the model learns patterns in a particular dataset without labels. In semi-supervised learning, some labels are present and others are not. In transfer learning, a model trained on a particular dataset is fine-tuned on another dataset, or with some of the parameters fixed. In delta learning, differences between data points are learned rather than absolute values. These differences can be, for example, between two computational chemistry calculations at different levels of theory. In reinforcement learning, a model is used to determine what action should be taken in a particular environment. A particular set of machine learning models is a neural network, which may include layers such as fully connected, convolutional, pooling, skip connections, etc. When many layers are connected to a neural network, it is called a deep learning model, which has many parameters and requires many data points to reduce the error in its predictions.

[0072] NISQ - Noisy Intermediate Scale Quantum Technology Any type of quantum computer may be suitable for the techniques disclosed herein. In accordance with the description herein, suitable quantum computers include, by way of non-limiting example, superconducting quantum computers (qubits implemented as small superconducting circuits - Josephson junctions) (Clarke, John, and Frank K. Wilhelm, "Superconducting quantum bits," Nature 453.7198 (2008):1031), trapped ion quantum computers (qubits implemented as ion-trapped states) (Kielpinski, David, Chris Monroe, and David J. Wineland, "Architecture for a large-scale ion-trap quantum computer." Nature 417.6890 (2002):709.), optical lattice quantum computers (qubits implemented as neutral atomic states trapped in an optical lattice) (Deutsch, Ivan H., Gavin K. Brennen, and Poul S. Jessen, "Quantum computing with neutral atoms in an optical lattice." Nature 417.6890 (2002):709.), and optical lattice quantum computers (qubits implemented as neutral atomic states trapped in an optical lattice). Quantum computing with neutral atoms in an optical lattice." arXiv preprint quant-ph / 0003022(2000)); spin-based quantum dot computers (qubits implemented as spin states of trapped electrons) (Imamog, A., David D. Awschalom, Guido Burkard, David P. DiVincenzo, Daniel Loss, M. Sherwin, and A. Small, "Quantum information processing using quantum dot spins and cavity QED.arXiv preprint quant-ph / 9904096 (1999)), spatially based quantum dot computers (qubits implemented as electron positions in double quantum dots) (Fedichkin, Leonid, Maxim Yanchenko, and KAValiev, "Novel coherent quantum bit using spatial quantization levels in semiconductor quantum dot." arXiv preprint quant-ph / 0006097 (2000)), coupled quantum wires (qubits implemented as pairs of quantum wires coupled by quantum point contacts) (Bertoni, A., Paolo Bordone, Rossella Brunetti, Carlo Jacoboni, and S. Reggiani, "Quantum logic gates based on coherent electron transport in quantum wires." Physical Review Letters 84, no. 25 (2000): 5912.), nuclear magnetic resonance quantum computers (qubits implemented as nuclear spins and probed by radio waves) (Cory, David G., Mark D. Price, and Timothy F. Havel, "Nuclear magnetic resonance spectroscopy: An experimentally accessible paradigm for quantum computing." arXiv preprint quant-ph / 9709001 (1997)), and solid-state NMR Kane-type quantum computers (qubits implemented as nuclear spin states of phosphorus donors in silicon) (Kane, Bruce E., "A silicon-based nuclear spin quantum computer.") Nature 393, no. 6681 (1998): 133.), electron-on-helium quantum computers (qubits implemented as electron spins) (Lyon, Stephen Aplin, "Spin-based quantum computing using electrons on liquid helium." arXiv preprint cond-mat / 0301581 (2006)), cavity quantum electrodynamics-based quantum computers (qubits implemented as states of trapped atoms coupled to high-finesse resonators) (Burell, Zachary, "An Introduction to Quantum Computing using Cavity QED concepts." arXiv preprint arXiv:1210.6512 (2012).), molecular magnet-based quantum computers (qubits implemented as spin states) (Leuenberger, Michael N., and Daniel Loss, "Quantum computing in Quantum computing in molecular magnets. arXiv preprint cond-mat / 0011415 (2001), fullerene-based electron spin resonance (ESR) quantum computers (qubits implemented as electron spins of atoms or molecules housed in fullerenes) (Harneit, Wolfgang, Quantum Computing with Endohedral Fullerenes. arXiv preprint arXiv:1708.09298 (2017)), linear optical quantum computers (qubits implemented as processing states of light in different modes through linear optical elements such as mirrors, beam splitters and phase shifters) (Knill, E., R. Laflamme, and G. Milburn, Efficient linear optics quantum computation.) by Nizovtsev, A.P., S. Ya Kilin, F. Jelezko, T. Gaebal, Iulian Popa, A. Gruber, and Jorg Wrachtrup, "A quantum computer based on NV centers in diamond: optically detected nutations of single electron and nuclear spins." Optics and spectroscopy 99, no. 2 (2005): 233-244.); Bose-Einstein condensate-based quantum computers (qubits implemented as two-component BECs) by Byrnes, Tim, Kai Wen, and Yoshihisa Yamamoto, "Macroscopic quantum computation using Bose-Einstein Macroscopic quantum computing with Bose-Einstein condensates." arXiv preprint quantum-ph / 1103.5512 (2011)); transistor-based quantum computers (qubits implemented as semiconductors coupled to nanophotonic cavities) (Sun, Shuo, Hyochul Kim, Zhouchen Luo, Glenn S. Solomon, and Edo Waks, "A single-photon switch and transistor enabled by a solid-state quantum memory.")," arXiv preprint quant-ph / 1805.01964 (2018)); quantum computers based on rare-earth metal ion-doped inorganic crystals (qubits implemented as hyperfine levels of the ground state of atoms in rare-earth ion-doped inorganic crystals) (Ohlsson, Nicklas, R. Krishna Mohan, and Stefan Kroll, "Quantum computer hardware based on rare-earth-ion-doped inorganic crystals," Optics Communications 201, no. 1-3 (2002): 71-77.); quantum computers based on metallic carbon nanospheres (qubits implemented as electron spins in conducting carbon nanospheres) (Nafradi, Balint, Mohammad Choucair, Klaus-Peter Dinse, and Laszlo Forro, "Room temperature manipulation of long lifetime spins in metallic-like carbon nanospheres" nanospheres." arXiv preprint cond-mat / 1611.07690 (2016)), as well as D-wave quantum annealers (qubits implemented as superconducting logic elements) (Johnson, Mark W., Mohammad HS Amin, Suzanne Gildert, Trevor Lanting, Firas Hamze, Neil Dickson, R. Harris, et al., "Quantum annealing with manufactured spins." Nature 473, no. 7346 (2011):194-198.), each of which is incorporated herein by reference in its entirety. In some cases, the quantum device is connected to the Internet to access the World Wide Web. In some cases, the quantum device is connected to a cloud computing infrastructure.In some cases, the quantum device is connected to an intranet. In some cases, the quantum device is connected to a data storage device.

[0073] Quantum Annealer A quantum annealer is an example of a quantum mechanical system that may consist of multiple qubits.

[0074] Each qubit is inductively coupled to a source of bias, called a local field bias. In some cases, the bias source is an electromagnetic device used to pass a magnetic flux through the qubit, thereby providing state control of the qubit (e.g., U.S. Patent Application Publication No. 2006 / 0225165, which is incorporated herein by reference in its entirety).

[0075] The local field biases on the qubits may be programmable and controllable. In some cases, a qubit control system including a digital processing device is connected to the system of qubits and is capable of programming and adjusting the local field biases on the qubits.

[0076] The quantum annealer may further include multiple couplings between multiple pairs of multiple qubits. In some cases, the coupling between two qubits is a device that is close to both qubits and passes magnetic flux to both qubits. In some cases, the coupling may include a superconducting circuit interrupted by a compound Josephson junction. The magnetic flux may pass through the compound Josephson junction, thereby passing magnetic flux to both qubits (e.g., U.S. Patent Application Publication No. 2006 / 0225165, which is incorporated by reference herein in its entirety). The strength of this magnetic flux may contribute quadratically to the energy of a quantum Ising model with a transverse magnetic field. In some cases, the coupling strength is enhanced by tuning a coupling device close to both qubits.

[0077] The coupling strength may be controllable and programmable. In some cases, a quantum annealer control system including a digital processing device may be connected to multiple couplings. In some cases, a quantum annealer control system including a digital processing device may be able to program the coupling strength of the quantum annealer.

[0078] In some cases, the quantum annealer performs a transformation of the quantum Ising model with a transverse magnetic field from an initial setup to a final setup. In some cases, the initial and final setups of the quantum Ising model with a transverse magnetic field provide a quantum system described by their corresponding initial and final Hamiltonians.

[0079] In some cases, quantum annealers may be used as heuristic optimizers for these energy functions. An example of such an analog processor is described in McGeoch, Catherine C. and Cong Wang (2013), "Experimental Evaluation of an Adiabatic Quantum System for Combinatorial Optimization," Computing Frontiers, May 14-16, 2013, and also in U.S. Patent Application No. 2006 / 0225165, each of which is incorporated herein by reference in its entirety.

[0080] In some cases, quantum annealing may be further used to provide samples from the Boltzmann distribution of the corresponding Ising model at finite temperature. For example, see "The Ising model: teaching an old problem new tricks" by Bian, Z., Chudak, F., Macready, WG, and Rose, G. (2010), and also "Quantum Boltzmann Machine" by Amin, MH, Andriyash, E., Rolfe, J., Kulchytskyy, B., and Melko, R. (2016), arXiv:1601.02036, which are incorporated herein by reference in their entirety. This sampling method is called quantum sampling.

[0081] In some cases, the quantum annealer is connected to the Internet to access the World Wide Web. In some cases, the quantum annealer is connected to a cloud computing infrastructure. In some cases, the quantum annealer is connected to an intranet. In some cases, the quantum annealer is connected to a data storage device.

[0082] High Performance Computing Devices The HPC may include one or more of a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), and a tensor streaming processor (TSP). Any other suitable processing unit capable of performing matrix multiplication may be used. Certain computing devices may be more efficient at operations such as matrix multiplication. These computing devices may provide additional efficiency improvements over other computing systems. For example, the matrix multiplication unit may be a GPU. A GPU may be a specialized electronic circuit optimized for high throughput, capable of performing the same set of operations in parallel on many blocks of data at once. For example, the matrix multiplication unit may be a TPU. A TPU may be a type of ASIC developed for low-bit precision processing by Google, see U.S. Patent Application Publication No. 2016 / 0342891 A1, the disclosure of which is incorporated by reference in its entirety. For example, the matrix multiplication unit may be an FPGA. An FPGA may be an integrated circuit chip that includes configurable logic blocks and programmable interconnects. An FPGA can be programmed after fabrication to execute custom algorithms. For example, the matrix multiplication device may be an ASIC. An ASIC may be an integrated circuit chip customized to execute a specific algorithm. In some cases, the ASIC cannot be programmed after manufacture. For example, the matrix multiplication device may be a TSP. A TSP may be a domain-specific programmable integrated chip designed for linear algebra calculations as they may be performed in artificial intelligence applications, an example of which may be described in "Groq rocks neural networks" by Gwennap, Linley, which is incorporated herein by reference in its entirety for all purposes. In some cases, the high-performance computing device is connected to the Internet to access the World Wide Web.In some cases, the high performance computing device is connected to a cloud computing infrastructure, in some cases, the high performance computing device is connected to an intranet, in some cases, the high performance computing device is connected to a data storage device.

[0083] Coherent Ising Machine The photonic computing device may be a coherent Ising machine (CIM). A CIM is a photonic device that is a powerful optimization system due to its enhanced, e.g., exhaustive, connectivity between spins. A coherent Ising machine may be adapted to solve the Ising problem, but instead, the light pulses may be used to represent continuous variables. Such a modification may also be referred to as a coherent optical network.

[0084] The coherent optical network may be of various types, such as any of the types described herein. In some cases, the coherent optical network includes optical computing devices. In some cases, the coherent optical network includes any or some of the optical instruments, such as beam splitters, free space lasers navigating on an optical bench, fiber ring cavities including optical fibers, and fiber optic instruments. In some cases, the coherent optical network includes integrated photonics.

[0085] In some cases, the coherent Ising machine computer is connected to the Internet to access the World Wide Web. In some cases, the coherent Ising machine computer is connected to a cloud computing infrastructure. In some cases, the coherent Ising machine computer is connected to an intranet. In some cases, the coherent Ising machine computer is connected to a data storage device.

[0086] Optical Computing Devices In some cases, the optical device includes a network of optical parametric oscillators (OPOs), as disclosed in U.S. Patent Application No. 2016 / 0162798 and International Application WO2015006494A1, the disclosures of each of which are incorporated by reference in their entirety.

[0087] In some cases, each spin in the Ising model is simulated by an optical parametric oscillator (OPO) operating at degeneracy.

[0088] A degenerate optical parametric oscillator (OPO) is an open dissipative system that undergoes a second-order phase transition at the oscillation threshold. Due to phase-sensitive amplification, a degenerate OPO may oscillate at a phase of either 0 or π with respect to the pump phase for amplitudes above the threshold. The phase is random and subject to quantum noise associated with optical parametric down-conversion during oscillation ramp-up. Thus, a degenerate OPO inherently represents a binary digit specified by its output phase. Based on this property, a degenerate OPO system may be used as a physical representative of an Ising spin system. The phase of each OPO is identified as an Ising spin, and its amplitude and phase are determined by the strength and sign of the Ising coupling between the associated spins.

[0089] When pumped by a strong source, a degenerate optical parametric oscillator (OPO) can adopt one of two phase states, corresponding to spin +1 or -1 in the Ising model. A network of N substantially identical OPOs with mutual coupling is pumped with the same source to simulate the Ising spin system. After a short period of time after the introduction of the pump, the network of OPOs approaches a steady state.

[0090] In some cases, such as when the amplitude is above threshold but the degenerate optical parametric oscillator is pumped by a weak source, it may represent a continuous variable instead of binary spin.

[0091] The phase state selection process relies on the mutual coupling of vacuum fluctuations and optical parametric oscillators (OPOs). In some cases, the pump is pulsed with a constant amplitude, in other implementations the pump power is gradually increased, and in still other implementations the pump is controlled in other ways.

[0092] In some examples of optical devices, the Ising model couplings are simulated by configurable couplings used to couple optical fields between optical parametric oscillators (OPOs). The configurable couplings may be configured to be off or configured to be on. The couplings may be switched on and off gradually or abruptly. When configured to be on, the configurations may provide any phase or amplitude depending on the Ising model coupling strength.

[0093] Each optical parametric oscillator (OPO) output may be interfered with by a phase reference, and the result may be captured by a photodetector. The OPO outputs represent configurations of an Ising model. For example, in the Ising model, zero phase may represent a spin--1 state, and π phase may represent a spin-+1 state.

[0094] For the Ising model with spin, according to some examples, the resonant cavities of multiple optical parametric oscillators (OPOs) are configured to have a round trip time equal to twice the period of the pulses from the pump source. Round trip time, as used herein, refers to the time it takes light to propagate along one path of the described recursive path. Pulses of a pulse train having a period equal to the period of the round trip time of the resonator cavities can propagate simultaneously through the OPOs without interfering with each other.

[0095] In some cases, optical parametric oscillator (OPO) coupling is provided by multiple delay lines distributed along the resonator cavity.

[0096] The multiple delay lines are equipped with multiple modulators that synchronously control the strength and phase of the coupling, allowing programming of the optical device to simulate the Ising model.

[0097] In a network of optical parametric oscillators (OPOs), a delay line and a corresponding modulator are sufficient to control the amplitude and phase of the coupling between every two OPOs.

[0098] In some cases, a device capable of sampling from the Ising model can be fabricated as a network of optical parametric oscillators (OPOs), as disclosed in U.S. Patent Application Publication No. 2016 / 0162798, the disclosure of which is incorporated by reference in its entirety.

[0099] In some cases, the network of optical parametric oscillators (OPOs) and the coupling of the OPOs are accomplished using optical elements such as commercially available mode-locked lasers and telecom fiber delay lines, modulators, and other optical devices. Alternatively, the network of OPOs and the coupling of the OPOs are implemented using fiber optic technology, such as fiber technology developed for telecommunications applications. In some cases, the coupling may be realized in fiber and controlled by an optical Kerr shutter.

[0100] In some cases, the optical computing device is connected to the Internet to access the World Wide Web, in some cases, the optical computing device is connected to a cloud computing infrastructure, in some cases, the optical computing device is connected to an intranet, in some cases, the optical computing device is connected to a data storage device.

[0101] Integrated photonic coherent Ising machine For example, an integrated photonic coherent Ising machine is disclosed in "Coherent Ising machines with error correction feedback" by Satoshi Kako, Timothee Leleu, Yoshitaka Inui, Farad Khoyratee, Sam Reifenstein, and Yoshihisa Yamamoto, which is incorporated by reference in its entirety, and is a combination of nodes and a connection network that solves a particular Ising problem. In some cases, the combination of nodes and a connection network may form an adiabatic optical computer. In other words, the combination of nodes and a connection network may solve an Ising problem non-deterministically when the values ​​stored in the nodes reach a steady state and minimize the energy of the nodes and the connection network. The values ​​stored in the nodes at the minimum energy level may be associated with values ​​that solve a particular Ising problem.

[0102] In some cases, the system includes a plurality of ring resonator photonic nodes, each of which stores a value; a pump coupled to each of the plurality of ring resonator photonic nodes via a pump waveguide for providing energy to each of the plurality of ring resonator photonic nodes; and a connection network comprising a plurality of 2×2 building blocks of elements, each element of the 2×2 building blocks comprising a plurality of phase shifters for adjusting the connection network with a parameter associated with an encoding of an Ising problem, wherein the connection network processes the value stored in each of the plurality of ring resonator photonic nodes, and the Ising problem is solved by the value stored in each of the plurality of ring resonator photonic nodes at a minimum energy level.

[0103] In some cases, the integrated photonic coherent Ising machine computer is connected to the Internet to access the World Wide Web. In some cases, the integrated photonic coherent Ising machine computer is connected to a cloud computing infrastructure. In some cases, the integrated photonic coherent Ising machine computer is connected to an intranet. In some cases, the integrated photonic coherent Ising machine computer is connected to a data storage device.

[0104] Digital Computer In some cases, the digital computer includes one or more hardware central processing units (CPUs) that perform the functions of the digital computer. In some cases, the digital computer further includes an operating system configured to execute the executable instructions. In some cases, the digital computer is connected to a computer network. In some cases, the digital computer is connected to the Internet to access the World Wide Web. In some cases, the digital computer is connected to a cloud computing infrastructure. In some cases, the digital computer is connected to an intranet. In some cases, the digital computer is connected to a data storage device.

[0105] Various types of digital computers may be used. Indeed, suitable digital computers may include, by way of non-limiting example, server computers, desktop computers, laptop computers, notebook computers, sub-notebook computers, netbook computers, set-top computers, media streaming devices, handheld computers, Internet appliances, portable smart phones, tablet computers, personal digital assistants, gaming consoles, and vehicles of communication. Smart phones may be suitable for use with one or more examples of the methods and systems described herein. In some cases, selected televisions, video players, and digital music players with computer network connectivity may be suitable for use in some instances of the systems and methods described herein. Suitable tablet computers may include tablet computers having booklet, slate, and convertible configurations.

[0106] In some cases, the digital computer includes an operating system configured to execute executable instructions. An operating system may be, for example, software including programs and data, that manages the device's hardware and provides services for the execution of applications. Various types of operating systems may be used. For example, suitable server operating systems include, by way of non-limiting examples, FreeBSD, OpenBSD, NetBSD, Linux, Apple® Mac OS X Server®, Oracle® Solaris®, Windows Server®, and Novell® NetWare®. Suitable personal computer operating systems may include, by way of non-limiting examples, UNIX-like operating systems such as Microsoft® Windows®, Apple® Mac OS X®, UNIX®, and GNU / Linux®. In some cases, the operating system is provided by cloud computing. Suitable mobile smartphone operating systems may include, by way of non-limiting example, Nokia® Symbian® OS, Apple® iOS®, Research In Motion® BlackBerry OS®, Google® Android®, Microsoft® Windows Phone® OS, Microsoft® Windows Mobile® OS, Linux®, and Palm® WebOS®.Suitable media streaming device operating systems may include, by non-limiting example, Apple TV®, Roku®, Boxee®, Google TV®, Google Chromecast®, Amazon Fire®, and Samsung® HomeSync®. Suitable game console operating systems may include, by non-limiting example, Sony® PS3®, Sony® PS4®, Microsoft® Xbox 360®, Microsoft® Xbox One®, Nintendo® Wii®, Nintendo® Wii U®, and Ouya®.

[0107] In some cases, the digital computer includes a storage and / or memory device. Various types of storage and / or memory may be used in a digital computer. In some cases, the storage and / or memory device includes one or more physical devices used to temporarily or permanently store data or programs. In some cases, the device includes a volatile memory and requires power to maintain the stored information. In some cases, the device includes a non-volatile memory and retains the stored information when the digital computer is not powered. In some cases, the non-volatile memory includes a flash memory. In some cases, the non-volatile memory includes a dynamic random access memory (DRAM). In some cases, the non-volatile memory includes a ferroelectric random access memory (FRAM). In some cases, the non-volatile memory includes a phase change random access memory (PRAM). In some cases, the device includes a storage device including, by non-limiting examples, a CD-ROM, a DVD, a flash memory device, a magnetic disk drive, a magnetic tape drive, an optical disk drive, and a cloud computing based storage. In some cases, the storage and / or memory device includes a combination of devices such as those disclosed herein.

[0108] In some cases, a digital computer includes a display used to provide visual information to a user. Various types of displays may be used. In some cases, the display includes a color cathode ray tube (CRT). In some cases, the display includes a liquid crystal display (LCD). In some cases, the display includes a thin film transistor liquid crystal display (TFT-LCD). In some cases, the display includes an organic light emitting diode (OLED) display. In some cases, the OLED display includes a passive matrix OLED (PMOLED) or an active matrix OLED (AMOLED) display. In some cases, the display includes a plasma display. In some cases, the display includes a video projector. In some cases, the display includes a combination of devices such as those disclosed herein.

[0109] In some cases, the digital computer includes an input device for receiving information from a user. Various types of input devices may be used. In some cases, the input device includes a keyboard. In some cases, the input device includes a pointing device, including, by non-limiting example, a mouse, a trackball, a trackpad, a joystick, a game controller, or a stylus. In some cases, the input device includes a touch screen or a multi-touch screen. In some cases, the input device includes a microphone for capturing voice or other audio input. In some cases, the input device includes a video camera or other sensor for capturing motion or visual input. In some cases, the input device includes a Kinect™, Leap Motion™, or the like. In some cases, the input device includes a combination of devices, such as those disclosed herein.

[0110]

[0013] Referring now to Figure 1, an example schematic of a system for testing molecules and their properties is shown. The system is configured to utilize machine learning (ML) models.

[0111] The system may include a digital computer 100. The digital computer may be various types of digital computers, such as, for example, digital computers described elsewhere herein. The digital computer may include at least one processing device 106 and at least one memory 112. The at least one memory may include a computer program executable by the processing device 106, which may be configured to obtain a representation of at least one property of a molecule and a request including a task, perform an inference against at least one machine learning (ML) model using the obtained representation, perform an inference confidence test, obtain a task result using a final result of the inference if the confidence is good, and obtain a task result by performing the obtained task if the confidence is bad, and electronically report the task result.

[0112] The system may include a computing platform 102 operatively connected to a digital computer 100. The computing platform 102 may include at least one processor 116. The at least one processor 116 may be various types of processors, such as, for example, types of processors described elsewhere herein. The at least one processor may include Noisy Intermediate Scale Quantum (NISQ) technology, such as those disclosed elsewhere herein, any quantum device, any high performance computing device, any quantum annealer, any optical computing device, an integrated photonic coherent Ising machine. For example, the at least one processor may include at least one field programmable gate array (FPGA), application specific integrated circuit (ASIC), central processing unit (CPU), graphics processing unit (GPU), tensor processing unit (TPU), tensor streaming processor (TSP), quantum computer, quantum annealer, integrated photonic coherent Ising machine, optical quantum computer, etc., or any combination thereof. The computing platform 102 may be provided by a cloud computing platform.

[0113] Each component of the system (e.g., hardware) may be used alone or in combination with other components (e.g., other hardware) as part of a system for performing the entire method, or any part thereof.

[0114] In some cases, the component may be used to obtain a representation of at least one property of a molecule and a request including a task, perform an inference against at least one machine learning (ML) model using the obtained representation, perform an inference confidence test, and if the confidence is good, obtain a task result using a final result of the inference, and if the confidence is bad, perform the task to obtain the task result.

[0115] The computing platform 102 may be operatively connected to the digital computer 100. The computing platform may include a readout control system 118. The readout control system may be configured to read information (e.g., calculation results, parameters, etc.) from the at least one processor 116. For example, the readout control system may be configured to convert data from the FPGA into data usable by the digital computer.

[0116] The system may include a database 104. The database 104 may be operatively connected to the digital computer 100. The database 104 may be various types of databases. The database 104 may refer to a central repository configured to store specifications of tasks and task results. In some cases, the database may be, for example, MongoDB. The database 104 may be used to store the representation of the molecular properties, the corresponding tasks, and their results. The database 104 may be used to store the task results. The database 104 may further be used to store the output from the chemical discovery toolkit. The database 104 may further be used to store a dataset for training the ML model. The dataset for training the ML model may be a subset or a complete set of the task results. The dataset for training the ML model may further be a subset or a complete set of the output from the chemical discovery toolkit. The processing device 106 may be further configured to store the representation of the molecular properties, the corresponding tasks, and their results in the database 104 and to retrieve the representation of the molecular properties from the database 104.

[0117] The computing platform 102 and the database 104 may be connected to the digital computer 100 over a network. The computing platform, the database, and / or the digital computer may have a network communication device. The network communication device may enable the computing platform, the database, and / or the digital computer to communicate with each other and with any number of user devices over a network. The network may be a wired or wireless network. For example, the network may be a fiber optic network, an Ethernet network, a satellite network, a cellular network, a Wi-Fi network, a Bluetooth network, etc. In one or more implementations, the computing platform, the database, and / or the digital computer may be several distributed computing platforms, databases, and / or digital computers accessible via the Internet. Such computing platforms, databases, and / or digital computers may be considered cloud computing devices. In some cases, one or more processors of the at least one processor may be located in the cloud.

[0118] At least one processor 116 may include one or more virtual machines. The one or more virtual machines may be one or more emulations of one or more computer systems. The virtual machines may be process virtual machines (e.g., a virtual machine configured to execute a process in a platform independent environment). The virtual machines may be system virtual machines (e.g., a virtual machine configured to execute an operating system and associated programs). The virtual machines may be configured to emulate a different architecture than the at least one processor. For example, the virtual machines may be configured to emulate a quantum computing architecture on a silicon computer chip. Examples of virtual machines include, but are not limited to, VMware®, VirtualBox®, Parallels®, QEMU®, Citrix® Hypervisor, Microsoft® Hyper-V®, and the like.

[0119] Referring now to FIG. 2, a flow diagram of an embodiment of a method for testing molecules and their properties is shown.

[0120] According to the process operations 200, an indication of at least one property of the molecule and a request including a corresponding task is obtained. The property of the molecule may be of various types, such as any of the properties described elsewhere herein. In one or more embodiments, the property of the molecule may include three-dimensional coordinates of atoms on the molecule, a fingerprint, a Simplified Molecular Input Line Entry System (SMILES), or an International Chemical Identifier (InChI).

[0121] In one or more embodiments, the molecular properties may further include ground state energy, excited state energy, highest occupied molecular orbital (HOMO)-lowest unoccupied molecular orbital (LUMO) gap, ionization potential, electron affinity, singlet-triplet gap, atomic charge, dipole moment, charge density, spectroscopic properties, peak position at X nm where X is the peak position, and binding affinity with a molecule of interest, equilibrium geometry, transition state geometry, reactivity, hydrophobicity, synthesizability, conformational entropy, and residence time of one molecule interacting with another molecule. In one or more embodiments, the molecular properties may further include effective carrier mass, acoustic wave propagation and elastic constants, band structure, density of states, and force and stress tensors on each atom. In one or more embodiments, the molecular properties may further include intercalation voltage, voltage profile, and phase diagram. In one or more embodiments, the molecular properties may further include radial distribution function, diffusion constant, viscosity, and electrical conductivity.

[0122] The corresponding task may be of various types, such as any of the tasks disclosed elsewhere herein. In one or more embodiments, the task may include at least one computational chemistry approach for calculating at least one property of a molecule as disclosed elsewhere herein. In one or more embodiments, the task may include prediction of a molecule having a desired property. In one or more embodiments, the tasks include at least one member of the group consisting of protein structure prediction, protein pocket identification, virtual screening based on protein-ligand interactions, quantitative structure-activity relationship prediction, molecular similarity search, classical force field generation, classical molecular dynamics simulation, binding free energy calculation, toxicity prediction, synthesizability prediction, generation of candidate catalyst molecules, conformational search, outcome prediction and barrier height based virtual screening, prediction of the reaction of catalyst candidates with substrates, ab initio molecular dynamics simulation, reaction barrier height calculation, molecular coordinate calculation, ground state energy calculation, excited state energy, HOMO-LUMO gap calculation, ionization potential calculation, electron affinity calculation, singlet-triplet gap calculation, atomic charge calculation, dipole moment calculation, charge density calculation, spectroscopic property prediction, equilibrium geometry prediction, reactivity prediction, hydrophobicity prediction, conformational entropy prediction, residence time prediction of a molecule interacting with another molecule.

[0123] In one or more embodiments, a task includes a pipeline of multiple subtasks.

[0124] In accordance with processing operation 202, the obtained representation is used to perform inference on at least one machine learning (ML) model. The at least one machine learning (ML) model is configured to mimic the outcome of the task. The at least one machine learning (ML) model may be of various types, such as any of the machine learning models (ML) described elsewhere herein. In one or more embodiments, the machine learning (ML) model is based on supervised learning. In one or more embodiments, the machine learning (ML) model is based on unsupervised learning. In one or more embodiments, the machine learning (ML) model is based on reinforcement learning. In one or more embodiments, the machine learning (ML) model is based on active learning. In one or more embodiments, the machine learning (ML) model is based on semi-supervised learning. In one or more embodiments, the machine learning (ML) model is based on continuous learning. In one or more embodiments, the machine learning (ML) model is based on transfer learning. In one or more embodiments, the machine learning (ML) model is based on delta learning. In one or more embodiments, modern machine learning (ML) models include artificial neural networks (ANNs), convolutional neural networks (CNNs), graph neural networks (GNNs), message passing neural networks, transformer networks, autoencoders (AEs), variational autoencoders (VAEs), and generative adversarial networks (GANs). These methods utilize automatic differentiation and gradient descent techniques. In one or more embodiments, classical machine learning (ML) models include kernel ridge regressors, random forest regressors, gradient boosting regressors, linear regressors, logistic regressors, ridge regressors, lasso regressors, polynomial regressors, Bayesian regressors, elastic net regressors, principal component regressors, least squares regressors, and support vector regressors. These classical ML models typically do not utilize automatic differentiation and use optimization strategies other than gradient descent. For all of the above models, ensemble models may be constructed by combining the predictions of two or more individual models.

[0125] Pursuant to processing operation 204, an inference reliability test is performed. In one or more embodiments, the inference reliability test includes a standard deviation calculation using the final result of the inference based on the multiple ML models. In one or more embodiments, the inference reliability test includes a standard error calculation using the final result of the inference based on the multiple ML models. In one or more embodiments, the inference reliability test includes a similarity search between the training dataset and the request to identify the rarity of the request. In some embodiments, the request may include an indication of at least one characteristic of the molecule and / or a corresponding task. For example, the request may include information required to compare a molecule with the rest of the training set (e.g., to identify the rarity of the request) and / or information required to pass into the ML model to obtain the result. In one or more embodiments, the inference reliability test includes an instance selection strategy for selecting a decisive instance that improves the ML model. The instance selection strategy includes an uncertainty sampling approach, a query by committee approach, a model changeability approach, an expected error reduction approach, and a density emphasis method. In one or more embodiments, accuracy requirements are obtained prior to processing act 202. In one or more embodiments, limitations on computational resources are obtained prior to processing act 202. The accuracy requirements and limitations on computational resources are used to perform an inference reliability test.

[0126] If the inference reliability test results are good, the final inference result is used to obtain a task result, according to process operation 206. In one or more embodiments, the final inference result along with the standard deviation based on multiple ML models as an indication of reliability will be electronically reported as the task result.

[0127] In accordance with process operation 208, if the result of the inference reliability test is poor, a task result is obtained by performing the task by using a representation of at least one property of the molecule and a corresponding task. In one or more embodiments, the task may be performed by using at least one computational chemistry approach in a chemical discovery toolkit. In one or more embodiments, the task may be performed by using at least one laboratory experiment approach in a chemical discovery toolkit. In one or more embodiments, the task includes a pipeline including multiple subtasks, the task is performed by performing all of the subtasks in a particular order.

[0128] In one or more embodiments, operation 208 may be performed in a chemical discovery toolkit. In one or more embodiments, the chemical discovery toolkit may include computational chemistry, quantum chemistry, and molecular mechanics methods. In one or more embodiments, the chemical discovery toolkit may further include a procedure for performing a task to predict at least one property of a molecule. In one or more embodiments, the chemical discovery toolkit may further include a pipeline for performing tasks performed in a particular order. In one or more embodiments, the tasks may include calculating energy, calculating electronic structure, optimizing molecular geometry, performing transition state search, performing conformation search, performing molecular similarity search, performing classical molecular dynamics simulation, performing ab initio molecular dynamics simulation, performing protein structure prediction, performing protein binding site prediction, performing virtual screening, performing protein-ligand binding structure prediction, performing free energy perturbation, performing ligand optimization, performing catalyst optimization, performing reaction path prediction, performing synthesizability prediction, performing spectroscopic information prediction, performing reactivity prediction, performing toxicity prediction, performing enzyme-substrate binding structure prediction, performing self-assembled nanomaterial structure prediction, optimizing material composition, and optimizing experimental conditions.

[0129] In one or more embodiments, calculating the electronic structure includes at least one member of the group consisting of Hartree-Fock (HF), density functional theory (DFT), coupled cluster single, double, and perturbative triple excitation (CCSD(T)), full configuration interaction (FCI), heat bath configuration interaction (HBCI), quantum Monte Carlo full configuration interaction (QMCFCI), density matrix embedding theory (DMET), fragment molecular orbital (FMO), incremental full configuration interaction (iFCI), Paulinet, hybrid quantum mechanics-molecular mechanics (QM / MM), and ML-based Schrödinger equation solvers such as ab initio molecular dynamics (AIMD) simulations, variational Monte Carlo, and diffusion Monte Carlo.

[0130] In one or more embodiments, the task results include at least one member of the group consisting of numbers indicating ground state energy, three-dimensional coordinates of atoms in the molecule, and numbers indicating peak positions of spectroscopic information.

[0131] In one or more embodiments, the obtained task results are used to train at least one ML model to improve its performance. In one or more embodiments, the obtained representation of at least one property of the molecule, the tasks, and the task results are stored in a database. The database may be of various types, such as any of the databases disclosed herein with respect to FIG. 1. In one or more embodiments, if the representation of at least one property of the molecule, the tasks, and the task results are stored in a database, prior to processing operation 202, the task results are retrieved from the database and electronically reported. In one or more embodiments, training may occur when the obtained task results stored in the database may lead to an improvement in the performance of the ML model. In one or more embodiments, training may occur when the number of obtained task results stored in the database is sufficient to improve the performance of the ML model. In one or more embodiments, training may occur based on a user request. In one or more embodiments, training may occur according to a predefined schedule.

[0132] In one or more embodiments, the ML model trained using at least one obtained task result may be uploaded to the system as an additional ML model in the system. In one or more embodiments, an existing ML model in the system may be replaced with the ML model trained using at least one obtained task result.

[0133] Following processing operation 210, task results are reported electronically. The results can be available via a user interface on a web page or returned to the user in a terminal window on a Unix system. The user can use the results to compare with results from traditional calculations or to compare the results with results from other ML models. The results can be used to perform molecular dynamics or to calculate binding energies, these are just a few examples.

[0134] Performing the task may include at least one pipeline consisting of generating a set of three-dimensional coordinates (conformations) of the molecule based on SMILES, performing geometry optimization calculations for each conformation by using one of the quantum chemistry methods described above, selecting a geometry at a local minimum by performing vibrational frequency calculations for each optimized geometry by using one of the quantum chemistry methods described above and ensuring there are no imaginary modes, and identifying the molecular conformation with the lowest energy as the optimized geometry of the requested molecule.

[0135] Performing the task may include at least one pipeline of performing reaction path prediction by using a Nudged Elastic Band (NEB) method and identifying an optimal reaction path between reactants and products, identifying a geometry with maximum energy along the reaction path as a transition state geometry, and obtaining a reaction barrier height as an energy difference between the energy of the reactants and the energy of the transition state geometry.

[0136] Performing the task may include at least one pipeline of performing excited state energy calculations for the molecule by using one of the quantum chemical methods described above, estimating oscillator strengths as a function of excitation energy (transition lines), and convolving these lines with a Gaussian function to form a spectrum.

[0137] Performing the task also includes at least one pipeline consisting of generating chemically plausible organic molecules, performing geometry optimization for each molecule, performing spectroscopic calculations for each molecule, and selecting and creating a list of molecules that have strong peaks at desired wavelengths.

[0138] Performing the task also includes at least one pipeline consisting of assigning classical force field parameters to each atom in the protein-ligand complex, adding water molecules and counterions as solvent to the protein-ligand complex, performing a geometry optimization to remove steric clashes between atoms in the system, performing a molecular dynamics simulation, measuring the simulation time that the ligand remains in the ligand binding site of the protein and reporting it as the residence time.

[0139] Performing the task may include at least one pipeline of downloading 3D coordinates of the protein from a Protein Data Bank (PDB) server by the requested PDB ID, adding missing atoms and residues to the protein coordinates, relaxing the protein structure by performing geometry optimization, identifying desired interaction sites between the protein and ligand by performing binding pocket prediction, generating 3D coordinates of the ligand based on SMILES in a given ligand database, performing binding free energy calculation between the ligand and the protein pocket for each ligand in the ligand database, sorting the ligands based on the strength of binding free energy, and identifying the ligand with maximum binding free energy.

[0140] Performing the task may include at least one experimental pipeline consisting of synthesizing Janus motifs by adding spacers between complementary guanine and cytosine pairs, examining the aggregation (nanotube formation) behavior for each Janus motif by using scanning electron microscopy (SEM), selecting Janus motifs that formed nanotubes, measuring the outer diameter of the nanotubes by performing tapping mode atomic force microscopy (TM-AFM) and transmission electron microscopy (TEM), and identifying Janus motifs that formed nanotubes with a desired outer diameter.

Claims

1. A method for testing molecules and their properties, comprising: a) obtaining a request including a representation of at least one characteristic of a molecule and a corresponding task; b) performing inference using the representation for at least one machine learning (ML) model, the at least one ML model being configured to mimic a result of the task to generate an inference outcome; c) performing an inference reliability test on the final result of the inference, i) obtaining a task result using the final result of said inference corresponding to a good inference reliability test; ii) performing the task using the representation in response to a failed inference reliability test to obtain a task result; d) outputting the task results; The method includes:

2. 2. The method of claim 1, further comprising: training the at least one ML model using a training dataset, the training dataset including the task results obtained in ii) of c) corresponding to the poor inference reliability test.

3. 2. The method of claim 1, further comprising storing in a database, in response to the failed inference reliability test, the representation of the at least one characteristic of the molecule, the corresponding task, and the task result obtained in ii) of c).

4. 4. The method of claim 3, further comprising, prior to b), outputting the representation of the at least one characteristic of the molecule, the corresponding task, and the task results if the task results are stored in the database.

5. The method of claim 1 , further comprising, prior to b), obtaining at least one of accuracy requirements and limitations on computational resources.

6. 2. The method of claim 1, wherein the representation of at least one property of a molecule comprises at least one member of the group consisting of three-dimensional coordinates of atoms on the molecule, a fingerprint, a Simplified Molecular Input Line Entry System (SMILES), an International Chemical Identifier (InChI), a spectroscopic geometry description, and a property name.

7. 2. The method of claim 1, wherein the at least one property comprises at least one member of the group consisting of molecular coordinates, ground state energy, excited state energy, highest occupied molecular orbital (HOMO)-lowest unoccupied molecular orbital (LUMO) gap, ionization potential, electron affinity, singlet-triplet gap, atomic charge, dipole moment, charge density, spectroscopic properties, peak position at X nm where X is the peak position, and binding affinity with a molecule of interest, equilibrium geometry, reactivity, hydrophobicity, synthesizability, conformational entropy, and residence time of a molecule interacting with another molecule.

8. 2. The method of claim 1, wherein the tasks include at least one member of the group consisting of protein structure prediction, protein pocket identification, virtual screening based on protein-ligand interactions, quantitative structure-activity relationship prediction, molecular similarity search, classical force field generation, classical molecular dynamics simulation, binding free energy calculation, toxicity prediction, synthesizability prediction, generation of candidate catalyst molecules, conformational search, virtual screening based on end result prediction and barrier height, prediction of the reaction of catalyst candidates with substrates, ab initio molecular dynamics simulation, reaction barrier height calculation, molecular coordinate calculation, ground state energy calculation, excited state energy, HOMO-LUMO gap calculation, ionization potential calculation, electron affinity calculation, singlet-triplet gap calculation, atomic charge calculation, dipole moment calculation, charge density calculation, spectroscopic property prediction, equilibrium geometry prediction, reactivity prediction, hydrophobicity prediction, conformational entropy prediction, residence time prediction of a molecule interacting with another molecule.

9. The method of claim 1 , wherein the task comprises a pipeline comprising a plurality of subtasks.

10. The method of claim 1 , wherein the task results include at least one member of the group consisting of numbers indicating ground state energy, three-dimensional coordinates of atoms in a molecule, and numbers indicating peak positions of spectroscopic information.

11. 2. The method of claim 1, wherein performing the tasks in ii) of c) comprises at least one member of the group consisting of: performing corresponding experiments, performing computational chemistry, quantum chemistry, molecular mechanics methods, calculating energy, calculating electronic structure, optimizing molecular geometry, performing transition state search, performing conformational search, performing molecular similarity search, performing classical molecular dynamics simulation, performing ab initio molecular dynamics simulation, performing protein structure prediction, performing protein binding site prediction, performing virtual screening, performing protein-ligand binding structure prediction, performing free energy perturbation, performing ligand optimization, performing catalyst optimization, performing reaction path prediction, performing synthesizability prediction, performing spectroscopic information prediction, performing reactivity prediction, performing toxicity prediction, performing enzyme-substrate binding structure prediction, performing self-assembled nanomaterial structure prediction, optimizing material composition, optimizing experimental conditions.

12. 12. The method of claim 11, wherein performing the structure prediction comprises using a quantum chemical method using at least one member of the group consisting of Hartree-Fock (HF), density functional theory (DFT), coupled cluster single, double, and perturbative triple excitation (CCSD(T)), full configuration interaction (FCI), heat bath configuration interaction (HBCI), quantum Monte Carlo full configuration interaction (QMCFCI), density matrix embedding theory (DMET), fragment molecular orbital (FMO), incremental full configuration interaction (iFCI), Paulinet, hybrid quantum mechanics-molecular mechanics (QM / MM), and ML-based Schrodinger equation solvers such as ab initio molecular dynamics (AIMD) simulations.

13. 2. The method of claim 1, wherein the at least one machine learning (ML) model comprises supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, active learning, delta learning, continual learning, and transfer learning.

14. The method of claim 1 , wherein the inference reliability test includes a standard deviation calculation using the final results of the inference.

15. The method of claim 1 , wherein the inference reliability test includes a standard error calculation using the final result of the inference.

16. The method of claim 1 , wherein the inference reliability test comprises a similarity search between a training data set and the requests to identify rarity of the requests.

17. 2. The method of claim 1, wherein the inference reliability test includes an instance selection strategy for selecting decisive instances that improve the ML model, the instance selection strategy including an uncertainty sampling approach, a query by committee approach, a model changeability approach, an expected error reduction approach, and a density emphasis method.

18. 1. A system for testing molecules and their properties, comprising: a) a memory component containing instructions for testing molecules and their properties, as well as at least one trained machine learning (ML) model; b) a computing platform operatively coupled to said memory component, said computing platform being operable to execute said instructions to perform at least i) obtaining a request including a representation of at least one property of a molecule and a corresponding task; ii) using the representation to perform inference on at least one machine learning (ML) model configured to generate an inference final result by mimicking the outcome of the task; iii) a computing platform configured to perform an inference reliability test on the final result of the inference, and to obtain a task result using the final result of the inference in response to a good inference reliability test, and to obtain a task result by performing the task using the representation in response to a bad inference reliability test; A system including:

19. 20. The system of claim 18, further comprising a database configured to store tasks and their results, and wherein the instructions include storing the representation of the at least one property of the molecule, the tasks, and the task results in the database.

20. 20. The system of claim 18, further comprising a distributed computing system.

21. 20. The system of claim 18, further comprising a cloud computing system.

22. 20. The system of claim 18, further comprising a classical computer.

23. 20. The system of claim 18, further comprising a non-classical computer.

24. 24. The system of claim 23, wherein the non-classical computer is a quantum computer.

25. 25. The system of claim 24, wherein the quantum computer comprises at least one member of the group consisting of a quantum device such as a superconducting quantum computer, a trapped ion quantum computer, an optical quantum computer, or a quantum annealer.

Images