Method and device for generating highly synthesizable molecules based on multi-precision feedback drive

Abstract

The present invention discloses a method and device for generating highly synthesizable molecules based on multi-precision feedback drive, comprising the following steps: constructing a hierarchical strategy network, wherein a reaction selection strategy network is used to select available reactions in the current state, and a reactant selection strategy network is used to select available reactants in the current state and perform chemical reactions according to reaction rules; constructing a multi-precision active learning framework, wherein a hierarchical strategy network is used to obtain generated molecules and selected precision labels and input them into a multi-precision substitution model, and each generated molecule is scored using the multi-precision substitution model, and an acquisition function is used to give a comprehensive score after considering the cost of using the scoring tool corresponding to the precision labels of different molecules and feedback to the hierarchical strategy network for gradient update. The present invention provides a more efficient, economical and reliable molecular design solution, accelerates the discovery and optimization process of new molecular entities, and promotes technological progress and innovative development in related fields.

Description

Technical Field

[0001] The present invention belongs to the technical field of molecule generation, and in particular relates to a method and device for generating highly synthesizable molecules based on multi-precision feedback drive. Background Art

[0002] Inspired by combinatorial chemistry, a number of molecular design methods have emerged in recent years that combine combinatorial chemistry with molecular generation. This involves assembling molecules in the form of chemical reactions, combining selectable reactants through constrained chemical reaction rules. Due to the unique properties of reaction-based assembly methods, they have gradually emerged from fragment-based methods and become a new basic method for molecular assembly (reaction-based methods). In recent years, several related algorithms have emerged, such as SynNet, PGFS, and SyntheMol. Among them, SyntheMol is currently the most cutting-edge and effective algorithm for generating molecules with high synthesizability.

[0003] SyntheMol uses the Monte Carlo Tree Search (MCTS) algorithm to identify high-scoring molecules by searching synthetic trees under the guidance of a property prediction model. During each MCTS expansion, SyntheMol constructs molecules by selecting building blocks and combining them through chemical reactions. The generated molecules are evaluated at their corresponding nodes to provide feedback to the MCTS algorithm. The node score includes four factors: molecular property prediction value P(N), development factor Q(N), exploration factor U(N), and building block diversity factor D(N). However, SyntheMol has three key problems:

[0004] 1) In the MCTS framework, each sample of generated molecules requires a complete rollout operation, that is, starting from the initial node, gradually expanding and evaluating each branch of the search tree until the complete molecule is constructed. This means that if 10,000 molecules are to be sampled, 10,000 independent rollout operations are required. Although this method can effectively explore the chemical space, the rollout process involves a large number of tree searches and node evaluations, which has a high computational cost and significantly reduces the sampling speed, especially when the chemical space is large and the synthesis path is complex. Therefore, when generating large-scale molecular libraries, the efficiency of MCTS sampling is limited by the computational burden of each rollout, resulting in a very slow overall sampling process. It is not suitable for real-world application scenarios that require large-scale sampling of millions of molecules. This problem also exists in the SynNet model;

[0005] 2) Given the complexity of molecular design (in the field of drugs and materials) and the long R&D cycles, it is unrealistic to rely on current generative models to find the so-called "optimal" molecules. Increasing the diversity of sampled molecules is like not putting all your eggs in one basket, which is an effective way to reduce R&D risks. However, existing generative algorithms do not have the advantage of sampling diversity from the underlying algorithm logic, and SyntheMol also has this problem.

[0006] 3) SyntheMol and all iterative feedback-based generative models require a scoring tool for each sample of generated molecules during training. Choosing a highly accurate scoring tool increases computational cost, inevitably leading to a significant increase in training time. This phenomenon is partly due to the high predictive accuracy of quantitative structure-activity relationship (QSAR) models currently constructed using deep learning methods, and their call cost is much lower than that of traditional docking scoring. Deep learning models typically perform well when sufficient representative data is available as training sets. However, in real-world scenarios, molecular design tasks often face the dilemma of sparse labeled data for small molecules. Active learning algorithms offer a promising solution to this data sparsity problem. The key lies in effectively sampling unlabeled data to maximize the prediction accuracy and generalization ability of the scoring model. Although various sampling strategies are currently available, these methods all rely on existing unlabeled data pools for sampling. Compared to the vast chemical space, the limited unlabeled data pool means that the structural diversity of the sampled molecules is limited, which also affects the generalization ability of the scoring model to a certain extent.

[0007] In summary, the molecular generation algorithm based on chemical reaction splicing still has obvious problems in sampling efficiency, sampling diversity and training cost that need to be solved urgently. Summary of the Invention

[0008] In view of the above, the purpose of the present invention is to provide a method and device for generating highly synthesizable molecules based on multi-precision feedback drive, so as to solve the problem of difficulty in implementing generation algorithms in molecular (drug / material field) design scenarios from the perspective of molecular synthesizability, improve the sampling diversity of generation algorithms and improve the exploration coverage in chemical space from the perspective of underlying algorithm logic, and reduce the cost of calling expensive high-precision scoring tools by generation models from the perspective of real application scenarios, so as to provide an efficient, economical and reliable molecular design solution for the fields of drug research and development and advanced materials development, accelerate the discovery and optimization process of new molecular entities, and promote technological progress and innovative development in related fields.

[0009] To achieve the above-mentioned purpose, the present invention provides the following technical solutions:

[0010] In a first aspect, an embodiment of the present invention provides a method for generating highly synthesizable molecules based on multi-precision feedback drive, comprising the following steps:

[0011] Highly synthesizable molecules are generated using a hierarchical strategy network comprising a reaction selection strategy network and a reactant selection strategy network, comprising: using the reaction selection strategy network to select an available reaction in a current state based on an initial molecular building block or an intermediate product in a previous state, using the reactant selection strategy network to select an available reactant in a current state based on the selected available reaction, and performing a chemical reaction according to a reaction rule of the selected available reaction to obtain an intermediate product or a final product;

[0012] A multi-precision active learning framework including a hierarchical strategy network, a Gaussian process as a multi-precision substitution model and a multi-precision maximum entropy search as an acquisition function is used to provide the required precision labels of the generated molecules while generating molecules, including: using the hierarchical strategy network to obtain the generated molecules and the selected precision labels and inputting them into the multi-precision substitution model, using the multi-precision substitution model to score each generated molecule, and using the acquisition function to give a comprehensive score after considering the usage cost of the scoring tool corresponding to the precision labels of different molecules and feeding it back to the reaction selection strategy network and the reactant selection strategy network for gradient update.

[0013] Preferably, the method further comprises:

[0014] Construct labeled data including reactants and their valid reaction labels, and use the labeled data, reactant dataset and reaction dataset to pre-train the reactant selection strategy network for multi-label classification tasks. The output classification results are the valid reactants and invalid reactants corresponding to each reaction, so that the reactant selection strategy network can pre-learn to identify optional reactants that match the selected available reactions under different reaction labels and pre-screen invalid reactants through pre-training.

[0015] Preferably, the molecular generation process is defined as a Markov decision process in which reactants react step by step based on reaction rules to obtain the final product molecules, and the generation flow network model is used to implement this process. The generation flow network defines the molecular generation process as a directed acyclic graph G = (S, A), where S represents the state space and A represents the action space. The vertex s0∈S is defined as the initial state with no incoming edges and corresponds to no initial molecule or a specific reactant. The vertex s is defined as f ∈S is the final state without outgoing edges and corresponds to the final generated molecule, defining the state sequence τ from the initial state to a certain terminal state = (s0→…→s n ) is a trajectory, and the molecular object x∈X realizes the forward state transfer s through sequential sampling actions t →s t+1, X represents the set of all molecules in the combinatorial chemical space, and the state transfer is to add new molecular fragments to the currently constructed molecules through chemical reactions;

[0016] In the hierarchical policy network, the state transfer s is based on the generation flow network t →s t+1 The action is divided into two steps. The first step is to select the available reactions in the current state based on the given reaction data set. The second step is to select available reactants based on the given reactant library and perform chemical reactions using the reaction rules of the available reactions selected in the first step. In the two steps, a reaction selection probability model is constructed. The reaction selection strategy network is used to select reactions and build a reactant selection probability model The reactant selection strategy network is used to select reactants, and the reaction selection strategy network and the reactant selection strategy network are trained to sample molecular objects with a probability π(x) proportional to the reward value R(x), that is, π(x)∝R(x) to generate diverse high-reward value molecular objects.

[0017] Preferably, the reaction selection strategy network is used to select available reactions in the current state based on the initial molecular building blocks or the intermediate products in the previous state, including:

[0018] When performing reaction selection, the initial molecular building blocks or intermediate products of the previous state are input into the reaction selection strategy network to obtain the original probability values ​​of all reactions. The classification probability distribution is constructed by shielding invalid reactions and normalizing the remaining valid reactions. Then, random sampling is performed from the classification probability distribution to obtain the available reactions in the current state.

[0019] Preferably, the reactant selection strategy network is used to select available reactants in the current state based on available reactions and to perform chemical reactions according to reaction rules of the available reactions to obtain intermediate products or final products, including:

[0020] When selecting reactants, the available reactions in the current state obtained by the reaction selection strategy network are input into the reactant selection strategy network to obtain the original probability values ​​of all reactants. The classification probability distribution is constructed by shielding invalid reactants and normalizing all remaining valid reactants. Then, random sampling is performed from the classification probability distribution to obtain the available reactants in the current state. Chemical reactions are carried out according to the reaction rules of the available reactions to obtain intermediate products or final products. After the chemical reaction is completed, the state is converted to the next state.

[0021] Preferably, the Gaussian process is modeled as a multi-precision surrogate model, wherein the kernel function adopts a multi-precision kernel.

[0022] In a second aspect, to achieve the above-mentioned object of the invention, an embodiment of the present invention further provides a highly synthesizable molecule generation device based on multi-precision feedback drive, which is implemented using the above-mentioned highly synthesizable molecule generation method based on multi-precision feedback drive, and includes: a hierarchical strategy network module and a multi-precision feedback drive module;

[0023] The hierarchical strategy network module is used to generate highly synthesizable molecules using a hierarchical strategy network including a reaction selection strategy network and a reactant selection strategy network, including: using the reaction selection strategy network to select an available reaction in the current state based on an initial molecular building block or an intermediate product in the previous state, using the reactant selection strategy network to select an available reactant in the current state based on the selected available reaction, and performing a chemical reaction according to the reaction rules of the selected available reaction to obtain an intermediate product or a final product;

[0024] The multi-precision feedback driving module is used to use a multi-precision active learning framework including a hierarchical strategy network, a Gaussian process as a multi-precision substitution model, and a multi-precision maximum entropy search as an acquisition function to provide the required precision labels of the generated molecules while generating molecules, including: using the hierarchical strategy network to obtain the generated molecules and the selected precision labels and inputting them into the multi-precision substitution model, using the multi-precision substitution model to score each generated molecule, using the acquisition function to give a comprehensive score after considering the usage cost of the scoring tool corresponding to the precision labels of different molecules and feeding it back to the reaction selection strategy network and the reactant selection strategy network for gradient update.

[0025] On the third aspect, in order to achieve the above-mentioned purpose of the invention, an embodiment of the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive when executing the computer program.

[0026] Fourthly, in order to achieve the above-mentioned purpose of the invention, an embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a computer, the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive is implemented.

[0027] In a fifth aspect, in order to achieve the above-mentioned purpose of the invention, an embodiment of the present invention further provides a computer product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) The present invention uses a generative flow network as a basis and an improved hierarchical strategy network obtained by combining the Markov decision process for molecule generation. The generative flow network algorithm is used to assign rewards corresponding to the generated molecules to each trajectory in the probability flow network, ensuring that the generation probability of different molecules is proportional to their reward values, thereby creating a distribution related to feedback and achieving independent and identically distributed sampling from it. Starting from the underlying algorithm logic, the diversity of molecular sampling is improved, and the exploration coverage in the chemical space is improved.

[0030] (2) The present invention introduces a pre-training strategy to pre-train the reactant selection strategy network, which can effectively solve the problem of excessive dimension of discrete action space, so that the reactant selection strategy network can focus more on establishing a reasonable probability distribution for selecting effective reactants, thereby improving computational efficiency.

[0031] (3) The present invention introduces a multi-precision active learning framework. Based on the active learning framework, Gaussian process regression is used as a multi-precision alternative model, and maximum entropy search is used as the acquisition function to train the hierarchical strategy network. Different scoring tools are selectively sampled by different molecules for scoring. While ensuring the overall scoring accuracy, the number of calls to high-precision scoring tools is reduced as much as possible, thereby reducing the overall training cost and ultimately achieving a balance between exploration, utilization, and cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0033] Figure 1 1 is a flow chart of a method for generating highly synthesizable molecules based on multi-precision feedback drive provided by an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the algorithm flow of the hierarchical strategy network provided by an embodiment of the present invention;

[0035] Figure 3 Schematic diagram of the algorithm flow of the multi-precision active learning framework provided by an embodiment of the present invention;

[0036] Figure 4 Schematic diagram of the algorithm framework of the hierarchical strategy network provided by an embodiment of the present invention;

[0037] Figure 5 It is a schematic structural diagram of a highly synthesizable molecule generation device based on multi-precision feedback drive provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0039] The inventive concept of the present invention is: in response to the problems of insufficient sampling efficiency and sampling diversity and high training cost of the molecular generation algorithm in the prior art, the embodiment of the present invention provides a highly synthesizable molecule generation method and device based on multi-precision feedback drive, constructs a hierarchical strategy network based on the generation flow network and the Markov decision process, and performs molecule generation through two steps of reaction selection and reactant selection, thereby ensuring that the generated molecules are synthesizable and improving the diversity of molecular sampling, further improving the computing efficiency through the pre-training strategy, and reducing the overall training cost by introducing a multi-precision active learning framework, providing an efficient, economical and reliable molecular design solution for the fields of drug research and development and advanced materials development.

[0040] Figure 1 FIG. 1 is a flow chart of a method for generating highly synthesizable molecules based on multi-precision feedback drive according to an embodiment of the present invention. Figure 1 As shown, the embodiment provides a method for generating highly synthesizable molecules based on multi-precision feedback drive, comprising the following steps:

[0041] S1. Using a hierarchical strategy network including a reaction selection strategy network and a reactant selection strategy network to generate highly synthesizable molecules, including: using the reaction selection strategy network to select an available reaction in the current state based on the initial molecular building blocks or the intermediate products in the previous state, using the reactant selection strategy network to select an available reactant in the current state based on the selected available reaction and performing a chemical reaction according to the reaction rules of the selected available reaction to obtain an intermediate product or a final product.

[0042] In the embodiment, the constructed hierarchical policy network is called SynGFN, which includes a reaction selection policy network (policy model 1) and a reactant selection policy network (policy model 2). SynGFN models molecular design as a series of cascade processes that simulate chemical reactions, gradually assembling novel molecules with synthesizable building blocks. SynGFN expresses goal-oriented molecular discovery as a Markov decision process that approximately satisfies a careful balance, thereby sampling a diverse set of high-performance molecules, each with a suggested synthetic route. During design, SynGFN faces two computational obstacles that cannot be ignored: (1) sampling from a probability distribution that is practically intractable; (2) evaluating reward functions (such as biological activity) is usually very expensive. SynGFN efficiently learns sampling by jointly adopting a custom pre-trained hierarchical policy network and uses a multi-precision active learning framework to strike a balance between accuracy and reward evaluation costs.

[0043] The underlying algorithm framework of SynGFN adopts a generative flow network (GFlowNet). Since the existing generative flow network does not incorporate reasonable physical and chemical constraints, it is difficult to ensure that the generated molecules are synthesizable. SynGFN models molecular generation as a Markov decision process, in which reactants react in the form of a synthetic tree through sequential reactions (i.e., reaction templates) and eventually reach the target molecule. In view of the complexity of assembling molecules through chemical reactions, SynGFN has designed a mechanism for performing hierarchical actions (decomposing each state transition in the Markov decision process into the selection of two sequential actions, first determining the reaction template and then selecting the corresponding appropriate reactants). The selection of hierarchical actions is guided by a hierarchical strategy network, that is, SynGFN contains strategy networks for reaction selection and reactant selection respectively. Chemical synthesis constraints are explicitly introduced in the design of SynGFN: during the generation process, only reactions and reactants that match the current state have a probability of being selected, and when invalid product molecules are generated, the reactions and reactants added in the current generation step will not ultimately be included in the generation process. For example Figure 2As shown in a, when selecting a reaction, the initial molecular building blocks or the intermediate products of the previous state are input into the reaction selection strategy network to obtain the original probability values ​​of all reactions. By shielding invalid reactions and normalizing the remaining valid reactions to construct a categorical probability distribution (specifically using a softmax distribution), random sampling is then performed from the categorical probability distribution. A temperature coefficient is introduced to adjust the probability distribution and regulate the randomness of the sampling to obtain the available reactions in the current state. Once a suitable reaction is selected, the reactant selection strategy network will be used to select reactants that are compatible with the subsequent reactions. When selecting reactants, the available reactions in the current state obtained by the reaction selection strategy network are input into the reactant selection strategy network to obtain the original probability values ​​of all reactants. By shielding invalid reactants and normalizing all remaining valid reactants to construct a categorical probability distribution, random sampling is then performed from the categorical probability distribution (specifically using a softmax distribution) to obtain the available reactants in the current state. Chemical reactions are then carried out according to the reaction rules of the available reactions to obtain intermediate products or final products. After the chemical reaction is completed, the state is converted to the next state. Without layering, the actions selected by the policy network need to specify both the reactant and the reaction based on the current state. A natural approach is to design actions as reactant + reaction labels, which introduces a large number of unnecessary dimensions. From a time complexity perspective, the time complexity of a single-layer action network is O(n1*n2), while the time complexity of a layered action network is O(n1+n2). Therefore, the design of a layered action network can greatly reduce computational complexity and thus improve computational efficiency.

[0044] What is even more challenging is that in the current task, the number of selectable reactants may be very large. Once a reaction is selected, the space of new selectable reactants will actually be updated. If a post-shielding operation is performed on the outputs of all reactants, the unnecessary information introduced by the extra dimensions will prevent the model from learning a reasonable probability distribution for selecting effective reactants. This will cause the high-dimensional discrete action space to be too large, making it difficult for the reactant selection strategy network to learn a reasonable probability distribution. To solve this problem, a pre-training strategy is designed for the reactant selection strategy network, and a multi-label classification task is designed to enable the reactant selection strategy network to pre-learn how to effectively identify selectable reactants that match the selected reaction under different reaction labels, thereby achieving a pre-shielding effect on invalid reactants. This allows the model to focus more on establishing a reasonable probability distribution for selecting effective reactants (such as Figure 2(as shown in b), construct labeled data including reactants and their valid reaction labels, use labeled data, reactant datasets and reaction datasets to pre-train the reactant selection strategy network for multi-label classification tasks, and output classification results for valid reactants and invalid reactants corresponding to each reaction. Combined with the multi-label classification loss, the reactant selection strategy network is optimized so that the reactant selection strategy network can pre-learn to identify optional reactants that match the selected available reactions under different reaction labels and pre-screen invalid reactants through pre-training.

[0045] Specifically, the molecular generation process is defined as a Markov decision process in which reactants react step by step based on reaction rules to obtain the final product molecules. The generation process of the generation flow network is defined as a directed acyclic graph G = (S, A), where S represents the state space and A represents the action space. The vertex s0∈S is defined as the initial state with no incoming edges. The corresponding initial molecule can be set to empty or to a specific reactant. In drug design, if there is a relatively good warhead structure for a certain target, the reactant containing the warhead substructure can be set as the common structure of all generated molecules, thereby flexibly transforming the problem from de novo molecule generation to a molecular optimization problem. In the embodiment, no initial molecule is uniformly used as the initial state. Define vertex s f ∈S is the final state without outgoing edges, corresponding to the final generated molecule. Define the state sequence τ from the initial state to a certain terminal state = (s0→…→s n ) is a trajectory, and the molecular object x∈X realizes the forward state transfer s through sequential sampling actions t →s t+1 , X represents the set of all molecules in the chemical space, and the state transition is to add new molecular fragments to the currently partially built molecules. Considering the complexity of the molecular assembly based on the reaction, the state transition s t →s t+1 The action is divided into two steps: the first step is to select the available reactions in the current state based on the given reaction data set; the second step is to select the appropriate reactants based on the given reactant library, and use the reaction rules selected in the first step to perform the chemical reaction. In the embodiment, SynGFN uniformly adopts two-step reaction as the maximum number of reaction steps, but the maximum number of reaction steps can be adjusted at will in actual application. Considering the selection of dual actions, it is necessary to construct the reaction selection probability model separately. Used to select reactions and build reactant selection probability models For reactant selection, SynGFN’s two policy models are trained to sample molecular objects with a probability τ(x) proportional to the reward value R(x), i.e., τ(x)∝R(x), to generate diverse molecular objects with high reward values.

[0046] The random forward policy PF (s t+1 |s t ) is decomposed and parameterized into two independent neural networks with learnable parameters and To fit the state transition s t →s t+1 distribution, θ1 and θ2 represent the learnable parameters of the reaction selection strategy network and the reactant selection strategy network, respectively. Similarly, the backward state transfer strategy can be parameterized as and The probability τ(x) of generating object x is based on P F It can be defined as:

[0047]

[0048] That is, the sum of the probabilities of all trajectories τ that map to the same object x. The goal of the generative flow network is to learn a strategy π such that the probability of sampling an object x∈X is proportional to its reward value R(x). Currently, there are several optional learning objectives for the generative flow network to ensure that π(x) ∝ R(x) during sampling. Among them, Trajectory Balance (TB) is the current mainstream choice, and its formula is as follows:

[0049]

[0050] Among them, Z θ Represents the partition function ∑ x∈X An approximate estimate of R(x), where n represents the trajectory length. TB loss Depending on the sampling of each trajectory, GFlowNet is trained based on off-policy trajectories, sampling trajectories τ~P F (τ;θ), and Perform a gradient descent step on it to make the state transition s in the trajectory τ t →s t+1 Based on the current design of two-step action to achieve state transition, the training objective is modified to:

[0051]

[0052] Furthermore, due to the characteristics of the current molecular generation order sampling, s t+1 Post-s t can be uniquely determined, so the reverse probability and All are fixed at 1, and the final training goal is:

[0053]

[0054] Using the final training goal Train the hierarchical policy network.

[0055] S2, using a multi-precision active learning framework including a hierarchical strategy network, a Gaussian process as a multi-precision substitution model and a multi-precision maximum entropy search as an acquisition function to provide the required precision labels of the generated molecules while generating molecules, including: using a hierarchical strategy network to obtain the generated molecules and the selected precision labels and input them into the multi-precision substitution model, using the multi-precision substitution model to score each generated molecule, using the acquisition function to give a comprehensive score after considering the usage cost of the scoring tool corresponding to the precision labels of different molecules and feeding it back to the reaction selection strategy network and the reactant selection strategy network for gradient update.

[0056] In the embodiment, a multi-precision active learning framework SynGFN-MFA is further developed based on SynGFN, which uses a multi-precision acquisition function to perform a sampling scheme of molecules during training (such as Figure 3 The multi-precision active learning framework is based on the active learning framework, using Gaussian process regression as the multi-precision surrogate model and maximum entropy search as the acquisition function to train SynGFN. High-scoring molecules are scored using computational tools of varying precision based on sampling, and labeled data is added to the dataset to update the multi-precision surrogate model, leading to iterative optimization of active learning.

[0057] Specifically, using Gaussian process as a multi-precision surrogate model, for a given series of data points z 1:n =(x1,m1),(x2,m2),…,(x n ,m n ) and the observed value y 1:n =y1,y2,…,y n , where m represents the precision label of the data and the observed value represents the score of the scoring tool. The objective function f is modeled as f|z using Gaussian process. 1:n ,y 1:n ~GP(μ n ,K n ), μ n Represents the mean, and the kernel function is the multi-precision kernel K n , GP(μ n ,K n ) represents a Gaussian process. For a general Gaussian process, we have:

[0058] μ n (x)=μ(x)+K(x,x 1:n )(K(x 1:n ,x 1:n )+σ 2 I) -1 (y 1:n -μ(x1:n ))#(5)

[0059] K n (x1,x2)=K(x1,x2)-K(x,x 1:n )(K(x 1:n ,x 1:n )+σ 2 I) -1 K(x 1:n ,x2)#(6)

[0060] Among them, μ n (x) represents the posterior mean of the new evaluation point x after a given series of n data points, μ(x) represents the prior mean at position x, that is, the initial predicted mean, K(·) represents the kernel function used to measure the similarity between data points, σ ​​represents the noise in the observation value, I represents the mutual information, μ(x) represents the prior mean at position x, that is, the initial predicted mean, K(·) represents the kernel function used to measure the similarity between data points, σ ​​represents the noise in the observation value, I represents the mutual information, and μ(x) represents the prior mean at position x. 1:n ) represents the mean vector based on the existing data points, K n (x1, x2) represents the posterior covariance function for new points x1 and x2 after a given series of n data points, K(x1, x2) represents the prior covariance function, x1 and x2 represent two independent new points to be evaluated, x 1:n Represents a collection of previously existing data points.

[0061] In the embodiment, for a new point to be evaluated z=(x, m), the kernel function uses a multi-precision kernel, which is expressed as:

[0062] K(z1,z2)=K1(x1,x2)×K2(m1,m2)#(7)

[0063] Here, K(z1,z2) represents a multi-precision kernel function, which is used to measure the similarity between two data points z1 and z2. In multi-precision modeling, the multi-precision kernel function considers both the input data point x and the corresponding precision information. K1(·) represents a square exponential kernel:

[0064]

[0065] Where l represents the length scale parameter, ‖x1-x2‖ 2 Represents the square of the Euclidean distance between x1-x2;

[0066] K2(·) represents the precision level kernel, which is used to measure the similarity between two precision levels m1 and m2:

[0067] K2(m1,m2)=c+(1-m1) 1+δ (1-m2) 1+δ #(9)

[0068] Among them, c,δ>0 are hyperparameters.

[0069] The acquisition function uses the Maximum Value Entropy Search (MES), and the formula is as follows:

[0070]

[0071] Where α(x,m) represents the acquisition function value of x when considering the accuracy level m. In a given dataset D j (j represents the number of active learning iterations) The multi-precision surrogate model output f at the evaluation point x at the accuracy level m m and the true objective function (here set to the highest precision M), the level of mutual information between is a regularization term used to balance the evaluation costs of different accuracy levels. Here, accuracy is assumed to be positively correlated with evaluation cost. Multi-precision MES aims to evaluate the amount of information that can be obtained from data tuples (x, m) outside the sampled dataset D. By selecting data points (x, m) that can maximize the mutual information I, it effectively guides the acquisition process. SynGFN is trained using the acquisition function as the reward function of the generative flow network, leveraging the diversity advantage of the generative flow network to provide high-scoring, diverse molecular objects. After training, the top K (Top K) data points (molecules and their precision labels) of the generated N data points (Sample N) are scored using the scoring method corresponding to the precision label and then added to the dataset D for the next round of active learning iterations.

[0072] In order to meet the requirements of the multi-precision active learning framework, SynGFN needs to be further modified. In the generative flow network algorithm, (x,m) is used as the sampling target object instead of just sampling the molecule itself, and the accuracy level of the scoring method is M. ′ ={0,1,2,…,M} is introduced as a new dimension into the state, and the state space changes from S to S M′ =S∪M′, in state transition s t →s t+1 Added additional action options for selecting the accuracy level.

[0073] In summary, the embodiment of the present invention provides a method for generating highly synthesizable molecules based on multi-precision feedback drive. In the constructed SynGFN, synthesis planning is deeply integrated into the molecular generation process. Compared with the standard generation flow network, a hierarchical strategy network is introduced to guide the reaction-based molecular assembly (first select the reaction template, and then select the reaction-compatible reactants) in the synthesis constraint molecule generation process under the generation flow network framework. By introducing the reaction-compatible reactant candidate mask and the unique pre-training technology of the strategy network, not only the computational complexity is reduced but also the convergence speed of the strategy network in the training stage is accelerated. In short, SynGFN directly generates reasonable synthesis paths (such as Figure 4 It connects the design and manufacturing stages that were previously separated in molecular design, and accelerates the design-manufacturing-testing-analysis (DMTA) cycle (as shown in Figure 1). Figure 4 (as shown in b in the figure). SynGFN uses a generative flow network algorithm to assign a reward corresponding to the generated molecule to each trajectory in the probabilistic flow network. This ensures that the generation probability of different molecules is proportional to their reward value (the target characteristic), thereby creating a distribution related to the feedback and enabling independent and identically distributed sampling from it. This approach addresses the problem of insufficient molecular diversity from a fundamental algorithmic perspective. Furthermore, a multi-precision active learning framework (called SynGFN-MFA) is constructed based on SynGFN, a multi-precision surrogate model, and an acquisition function. This framework reduces the cost of scoring and evaluation while maintaining the quality of the molecule generation itself.

[0074] The following verification experiments are designed based on the effects of SynGFN and SynGFN-MFA provided in the embodiments of the present invention to fully illustrate the effectiveness of the solution.

[0075] (1) The pre-training strategy introduced by SynGFN can effectively solve the problem of excessive dimension of discrete action space and improve computational efficiency.

[0076] Using soluble epoxide hydrolase (sEH) as a drug target (i.e., using the binding affinity of generated molecules to the sEH target as the reward source), the training performance of SynGFN with and without pre-training was compared for four different reactant library sizes (1,000, 10,000, 30,000, and 100,000 molecules). Pre-training was found to significantly improve SynGFN's exploration efficiency when the reactant library size reached 100,000 (i.e., the output dimension of the reactant selection strategy network was 100,000), enabling it to more rapidly and effectively explore more reasonable regions of high-dimensional chemical space. Specifically, the average reward score of sampled molecules per training iteration and the average reward score of the top 10 sampled molecules were used as monitoring metrics. Upon convergence, SynGFN using pre-training achieved improvements of 37.5% and 15.6%, respectively, in these two metrics.

[0077] (2) SynGFN has excellent spatial exploration efficiency.

[0078] To compare SynGFN with traditional virtual screening methods, a virtual chemical space constructed based on a mini-reactant library (1,879 molecules) was enumerated, ultimately yielding a chemical space with 300 million compounds. For SynGFN, its training model involves traversing the molecules in this chemical space and learning a reasonable probability distribution for actions based on reward feedback. To demonstrate SynGFN's spatial search advantage, a random selection of 1 million compounds, 10 times the number of molecules traversed during SynGFN training, was used as a virtual screening compound library for comparison. Taking the sEH target as an example, under the same QSAR activity prediction model, the top 10,000 molecules searched during SynGFN training had a higher coverage of the chemical space than the top 10,000 molecules after virtual screening of 1 million. Only one-tenth of the number of molecules needs to be screened to effectively identify and discover the spatial region of highly active molecules. Naturally, it can be seen that the score distribution of molecules traversed by SynGFN is also better than that obtained through virtual screening alone. It should be noted that for SynGFN, only about 100,000 molecules need to be traversed to complete training (mean reward reaches a stable value), which means that there is no need to continue training in the future, and a sufficiently diverse range of high-scoring molecules can be directly sampled. The search performance of SynGFN and virtual screening was further compared within different QSAR activity score thresholds (here, since the comparison is the number of high-scoring compounds screened, in order to ensure fairness, virtual screening uses a virtual library with the same number of molecules as SynGFN search). Using the enrichment factor as an indicator (the ratio of the number of high-scoring compounds searched by SynGFN to the number of high-scoring compounds searched by virtual screening), it can be more intuitively seen that as the score threshold increases, the number of high-scoring compounds searched by SynGFN has increased by dozens of times. The results show that SynGFN can explore more high-scoring compounds.

[0079] (3) SynGFN is significantly better than SyntheMol in sampling molecular diversity (i.e., chemical space exploration).

[0080] The diversity of generated molecules essentially reflects the extent to which a generative model explores its chemical space. #Circles is used as a metric to evaluate the chemical space coverage of different generative models. #Circles is inspired by the concept of packing number in topology. The packing number refers to the maximum number of non-overlapping spheres that can be used to fill a given volume. Given a distance threshold t, #Circles calculates the maximum number of non-overlapping circles that can fit in chemical space. By determining this maximum set of non-overlapping molecules, this metric effectively assesses chemical space coverage. This assessment method helps understand the depth and breadth of a generative model's exploration of chemical space, and thus evaluates its ability to generate diverse molecules. Using #Circles as a metric provides a quantitative and comparable indicator that helps identify generative models that perform better in terms of exploration and coverage of chemical space. Aurora kinase A (AURKA), D2 dopamine receptor (DRD2), and soluble epoxide hydrolase (sEH) were selected as therapeutic targets for benchmarking the drug design quality of various approaches. These three targets are currently very popular in the treatment of cancer, schizophrenia, Parkinson's disease, and Alzheimer's disease, respectively. We first examined the #Circles curves of molecules generated by different models (SynGFN and SyntheMol) at different distance thresholds t (10,000 sampled molecules and the top 1,000 scoring molecules). SynGFN consistently exhibited the best #Circles values ​​across different thresholds and targets, particularly in the mid-to-high threshold range (t>0.65), demonstrating its superior global spatial coverage. We further summarized the performance of the different generative models at three threshold settings: t = 0.7, 0.75, and 0.8. The results showed that SynGFN demonstrated superior chemical space coverage in the search for active molecules, targeting these three different targets. Notably, compared to the baseline model SyntheMol, SynGFN achieved up to 70-fold improvement in chemical space coverage for the top 1,000 scoring generated molecules. For multiple targets and evaluation settings, SynGFN-XL achieved improvements ranging from 2 to 13-fold for the sampled 10,000 molecules. We also examined the distribution of diversity-adjusted QSAR scores for the generated molecules, obtained by clustering the 10,000 sampled molecules and the top 1,000 subset molecules into #Circles, retaining the highest-scoring molecules from each circle. The top 1,000 molecules generated by SynGFN at all scales exhibited superior diversity-adjusted QSAR scores for all three targets. This highlights SynGFN's superior ability to balance exploration and exploitation while searching for ideal molecules in chemical space.

[0081] (4) SynGFN-MFA can effectively reduce the annotation cost of the drug discovery process.

[0082] Using sEH as a target case study, we examined the effectiveness of multi-fidelity active-learning (MF-AL) at a mini-reactor library scale (1879 molecules) to demonstrate the algorithm's performance. Here, we used a QSAR activity prediction model for sEH (based on a message-passing neural network (MPNN)) as a low-precision scoring method and AutoDock Vina as a high-precision scoring method. To simulate the practical challenges faced in real-world drug design, where activity data are scarce and high-precision scoring is expensive, we set the initial dataset size to 1000, and examined the amount of annotated data in each active learning round, ranging from 50 to 200. We compared the performance of SynGFN using MF-AL with that using only the high-precision scoring method (herein referred to as SF-AL). Specifically, we examined the docking scores of the top 100 molecules sampled from a sample of 1000 molecules after each active learning round. To account for the time difference between QSAR activity scoring and docking scoring, we set the time cost of the two to be 1:100. The final result is that under the multi-precision active learning framework, SynGFN can achieve the same effect (top-100 score on par with SF-AL) at a lower cost. Under different annotated data sizes (the amount of annotated data per round is 50 / 100 / 150 / 200 respectively), MF-AL can reduce costs by 30%-80%. This advantage becomes more obvious as the amount of annotated data decreases. When the amount of annotated data per round is 50, MF-AL can reduce costs by nearly 80%, which is very attractive for drug design in real-world scenarios.

[0083] (5) SynGFN has demonstrated its effectiveness and significant advantages in real drug molecule scenarios (the designed synthesis pathway is nearly 100% consistent with the actual synthesis pathway).

[0084] Based on SynGFN's promising performance in a series of previous tests, it was applied to the design of NMDAR inhibitors. Historically, conventional NMDARs have been one of the most sought-after drug development targets for neurological diseases due to their critical role in neuronal development and synaptic plasticity in the central nervous system. Compositionally, NMDARs are composed of a diverse set of receptor subunits. Classical NMDA receptors typically form a heterotetramer consisting of two GluN1 and two GluN2 subunits. GluN1 / GluN3A is also an NMDA receptor subtype, but as excitatory glycine receptors (eGlyRs), they differ significantly from conventional NMDARs and are considered an NMDAR-related receptor subfamily awaiting further investigation. eGlyRs have been revealed to be bona fide brain receptors, functionally expressed in multiple brain regions, making them potential novel therapeutic targets for neuropsychiatric disorders. However, eGlyR biology remains in its infancy and has yet to be widely explored as a target. A major obstacle is the lack of a crystal structure for GluN1 / GluN3A (due to the lack of a full-length GluN3A receptor structure). At the same time, current protein structure prediction tools have significant errors in predicting the active conformations of multi-subunit proteins. In this context, structure-based drug design approaches face significant uncertainty. Several compounds have been developed that antagonize GluN1 / GIuN3A. Due to the high structural similarity between the orthosteric ligand-binding pockets of GluN1 and GluN3A, designing compounds that selectively target the orthosteric site of GluN3A is challenging. A more promising approach is to design negative allosteric modulators (NAMs), such as EU1180-438 and WZB117, which are thought to act by binding to the TMD site near the GluN3A pre-M1 region. Therefore, we aim to design novel GluN1 / GIuN3A inhibitors by leveraging SynGFN's ability to efficiently and accurately explore chemical space, starting with a ligand-based drug design approach. The 3D conformations of WZB117 and EU1180-438 molecules were used as shape queries for ROCS. The generated molecules were constructed using OMEGA, and their similarity to the query was calculated using ROCS, a 3D shape superposition method, as a reward score to guide the generation process. Ultimately, 10 candidate compounds were selected for chemical synthesis and biological activity verification. All compounds were successfully synthesized within 30 days, with six demonstrating moderate inhibitory activity against GluN1 / GluN3A. The most potent compound, D1, had an IC50 value of 2.68 μM. 60% of the tested molecules showed potential activity, and the synthetic routes provided by SynGFN were nearly 100% consistent with the actual synthetic routes.In the past, drug molecules converted using generative AI technology often needed to be screened individually by synthesis experts during actual screening. Each molecule required tedious synthesis route confirmation (including reactants, catalyst prices, reaction conditions, post-reaction treatment methods, etc.). Based on current testing alone, a full-time equivalent (FTE) can only synthesize 2-3 compounds per month. With the help of SynGFN, this efficiency has increased by 100%. With the help of SynGFN, this process can be greatly accelerated. SynGFN connects the design and synthesis links, breaking through the bottleneck of the slow DMTA cycle.

[0085] Based on the same inventive concept, Figure 5 As shown, an embodiment of the present invention further provides a highly synthesizable molecule generation device 500 based on multi-precision feedback drive, comprising: a hierarchical strategy network module 510 and a multi-precision feedback drive module 520 .

[0086] Hierarchical strategy network module 510 is used to generate highly synthesizable molecules using a hierarchical strategy network comprising a reaction selection strategy network and a reactant selection strategy network. This includes: using the reaction selection strategy network to select available reactions in the current state based on initial molecular building blocks or intermediate products in the previous state; using the reactant selection strategy network to select available reactants in the current state based on the available reactions; and performing chemical reactions according to the reaction rules of the available reactions to obtain intermediate products or final products.

[0087] The multi-precision feedback driving module 520 is used to optimize the reactant selection in the reactant selection strategy network using a multi-precision active learning framework including a hierarchical strategy network, a multi-precision substitution model and an acquisition function, including: using the reactant selection strategy network in the hierarchical strategy network to obtain optional reactants and the precision label of each optional reactant and input them into the multi-precision substitution model, using the multi-precision substitution model to select different scoring tools according to different precision labels to score each optional reactant, and using the acquisition function to sample the available reactants according to the scoring results and feed them back to the reactant selection strategy network.

[0088] Based on the same inventive concept, an embodiment of the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive when executing the computer program.

[0089] Based on the same inventive concept, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a computer, the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive is implemented.

[0090] Based on the same inventive concept, an embodiment of the present invention further provides a computer product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for generating highly synthesizable molecules based on multi-precision feedback drive.

[0091] It should be noted that the highly synthesizable molecule generation device based on multi-precision feedback drive, electronic device, computer-readable storage medium and computer product provided in the above embodiments all belong to the same inventive concept as the highly synthesizable molecule generation method based on multi-precision feedback drive. The specific implementation process is detailed in the embodiment of the highly synthesizable molecule generation method based on multi-precision feedback drive, which will not be repeated here.

[0092] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating highly synthesizable molecules based on multi-precision feedback drive, characterized in that: The following steps are involved: Highly synthesizable molecules are generated using a hierarchical strategy network comprising a reaction selection strategy network and a reactant selection strategy network, comprising: using the reaction selection strategy network to select an available reaction in a current state based on an initial molecular building block or an intermediate product in a previous state, using the reactant selection strategy network to select an available reactant in a current state based on the selected available reaction and performing a chemical reaction according to a reaction rule of the selected reaction to obtain an intermediate product or a final product; The molecular generation process is defined as a Markov decision process in which reactants react step by step based on reaction rules to obtain the final product molecules, and the generation flow network model is used to implement this process. The generation flow network defines the molecular generation process as a directed acyclic graph G = (S, A), where S represents the state space and A represents the action space. The vertex s0∈S is defined as the initial state with no incoming edges and corresponds to no initial molecules or a specific reactant. The vertex s is defined as f ∈S is the final state without outgoing edges and corresponds to the final generated molecule, defining the state sequence τ from the initial state to a certain terminal state = (s0→…→s n ) is a trajectory, and the molecular object x∈X realizes the forward state transfer s through sequential sampling actions t →s t+1 , X represents the set of all molecules in the combinatorial chemical space, and the state transfer is to add new molecular fragments to the currently built molecules through chemical reactions; in the hierarchical strategy network, the state transfer s is based on the generation flow network. t →s t+1 The action is divided into two steps. The first step is to select the available reactions in the current state based on the given reaction data set. The second step is to select available reactants based on the given reactant library and perform chemical reactions using the reaction rules of the available reactions selected in the first step. In the two steps, a reaction selection probability model is constructed. The reaction selection strategy network is used to select reactions and build a reactant selection probability model The reactant selection strategy network is used to select reactants, and the reaction selection strategy network and the reactant selection strategy network are trained to sample molecular objects with a probability π(x) proportional to the reward value R(x), that is, π(x)∝R(x), to generate diverse high-reward molecular objects. A multi-precision active learning framework including a hierarchical strategy network, a Gaussian process as a multi-precision surrogate model, and a multi-precision maximum entropy search as an acquisition function is used to provide the required precision labels of the generated molecules while generating molecules, including: using the hierarchical strategy network to obtain the generated molecules and the selected precision labels and inputting them into the multi-precision surrogate model, using the multi-precision surrogate model to score each generated molecule, and using the acquisition function to give a comprehensive score after considering the usage cost of the scoring tool corresponding to the precision labels of different molecules and feeding it back to the reaction selection strategy network and the reactant selection strategy network for gradient update.

2. The method for generating highly synthesizable molecules based on multi-precision feedback drive according to claim 1, characterized in that: The method further comprises: Construct labeled data including reactants and their valid reaction labels, and use the labeled data, reactant dataset and reaction dataset to pre-train the reactant selection strategy network for multi-label classification tasks. The output classification results are the valid reactants and invalid reactants corresponding to each reaction, so that the reactant selection strategy network can pre-learn to identify optional reactants that match the selected available reactions under different reaction labels and pre-screen invalid reactants through pre-training.

3. The method for generating highly synthesizable molecules based on multi-precision feedback drive according to claim 1, characterized in that: The reaction selection strategy network is used to select available reactions in the current state according to the initial molecular building blocks or the intermediate products in the previous state, including: When performing reaction selection, the initial molecular building blocks or intermediate products of the previous state are input into the reaction selection strategy network to obtain the original probability values ​​of all reactions. The classification probability distribution is constructed by shielding invalid reactions and normalizing the remaining valid reactions. Then, random sampling is performed from the classification probability distribution to obtain the available reactions in the current state.

4. The method for generating highly synthesizable molecules based on multi-precision feedback drive according to claim 1, characterized in that: The reactant selection strategy network is used to select available reactants in the current state based on the selected available reactions and to perform chemical reactions according to the reaction rules of the selected available reactions to obtain intermediate products or final products, including: When selecting reactants, the available reactions in the current state obtained by the reaction selection strategy network are input into the reactant selection strategy network to obtain the original probability values ​​of all reactants. The classification probability distribution is constructed by shielding invalid reactants and normalizing all remaining valid reactants. Then, random sampling is performed from the classification probability distribution to obtain the available reactants in the current state. Chemical reactions are carried out according to the reaction rules of the available reactions to obtain intermediate products or final products. After the chemical reaction is completed, the state is converted to the next state.

5. The method for generating highly synthesizable molecules based on multi-precision feedback drive according to claim 1, characterized in that: The Gaussian process is modeled as a multi-precision surrogate model, where the kernel function adopts a multi-precision kernel.

6. A highly synthesizable molecule generation device based on multi-precision feedback drive, implemented using the highly synthesizable molecule generation method based on multi-precision feedback drive according to any one of claims 1 to 5, characterized in that: include: Hierarchical strategy network module and multi-precision feedback drive module; The hierarchical strategy network module is used to generate highly synthesizable molecules using a hierarchical strategy network including a reaction selection strategy network and a reactant selection strategy network, including: using the reaction selection strategy network to select an available reaction in the current state based on an initial molecular building block or an intermediate product in the previous state, using the reactant selection strategy network to select an available reactant in the current state based on the selected available reaction, and performing a chemical reaction according to the reaction rules of the selected available reaction to obtain an intermediate product or a final product; The multi-precision feedback driving module is used to use a multi-precision active learning framework including a hierarchical strategy network, a Gaussian process as a multi-precision substitution model, and a multi-precision maximum entropy search as an acquisition function to provide the required precision labels of the generated molecules while generating molecules, including: using the hierarchical strategy network to obtain the generated molecules and the selected precision labels and inputting them into the multi-precision substitution model, using the multi-precision substitution model to score each generated molecule, using the acquisition function to give a comprehensive score after considering the usage cost of the scoring tool corresponding to the precision labels of different molecules and feeding it back to the reaction selection strategy network and the reactant selection strategy network for gradient update.

7. An electronic device comprising a memory and a processor, wherein the memory is used to store a computer program, wherein: The processor is configured to implement the method for generating highly synthesizable molecules based on multi-precision feedback drive according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a computer, the method for generating highly synthesizable molecules based on multi-precision feedback drive according to any one of claims 1 to 5 is implemented.

9. A computer product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for generating highly synthesizable molecules based on multi-precision feedback drive according to any one of claims 1 to 5 is implemented.

Images