Condition screening method based on literature data, and electronic device

Abstract

The present application discloses a condition screening method based on literature data, and an electronic device. The method comprises: acquiring reaction conditions and reactants in literature data, traversing the reaction conditions, and matching the reactant covered by the current reaction condition; initializing a condition set and the coverage thereof; performing dimensionality reduction transformation on reactants covered by all the conditions into three-dimensional points; calculating the effect on the coverage of the condition set after a single reaction condition is added to the condition set, selecting a reaction condition having the largest effect, adding same to the condition set, updating the condition set and the coverage thereof, and iterating until the number of reaction conditions added reaches a set value or all the reaction conditions are added to the condition set; and outputting the condition set. The condition screening method and the electronic device provided in the present application avoid the problem in the related art of time-consuming and labor-intensive manual operations, and successfully apply the computer technology to the screening of reaction conditions in literature data, thereby facilitating the automation and intelligence of chemical development.

Description

A conditional screening method and electronic device based on literature data

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202410795122.8, filed on June 19, 2024, entitled "A Conditional Screening Method and Electronic Device Based on Document Data", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of chemical reaction condition screening technology, and in particular to a condition screening method and electronic device based on literature data. Background Technology

[0004] In chemical research and industrial applications, high-throughput chemistry utilizes automated equipment and high-throughput technologies to rapidly and efficiently conduct large-scale chemical reactions and compound synthesis. Optimizing chemical reaction conditions plays a crucial role in improving reaction efficiency, reducing energy consumption, minimizing byproduct formation, and enhancing product purity. However, traditional methods for screening reaction conditions typically rely on manual literature review, manual experiments, and tedious data analysis. This is not only time-consuming and labor-intensive but also fails to comprehensively cover all possible combinations of reaction conditions, thus limiting the optimization potential of chemical reactions.

[0005] With the development of computer technology, more and more computer technology is being used in the field of chemistry, showing characteristics such as low cost and high efficiency. However, there are no reports on how to use computer technology to screen reaction conditions from literature.

[0006] In summary, this application provides a conditional screening method and electronic device based on literature data, which can be applied to the screening and processing of data in chemical reactions, especially in high-throughput chemistry, thereby solving the above-mentioned problems.

[0007] Summary of the Invention

[0008] To address the time-consuming and labor-intensive nature of manual operations in related technologies, this application discloses a conditional screening method based on literature data, the specific technical solution of which is as follows:

[0009] A conditional screening method based on literature data includes:

[0010] Obtain the reaction conditions and reactants from the literature data, iterate through each reaction condition, and match the reactants covered by the current reaction condition.

[0011] Initialize the condition set and its coverage;

[0012] All reactants covered by all conditions are reduced in dimension and converted into three-dimensional points;

[0013] Calculate the impact of adding a single reaction condition to the condition set on the condition set coverage, select the reaction condition with the greatest impact to add to the condition set, update the condition set and its coverage, and iterate until the number of reaction conditions added reaches the set value or all reaction conditions have been added to the condition set.

[0014] Output the condition set.

[0015] Optionally, the reaction conditions include information such as the catalyst, ligand, solvent, amount, and temperature used in the reaction.

[0016] Optionally, the reaction conditions and reactants obtained from literature data are unique, and duplicate reaction conditions and reactants will be deleted. However, when matching reaction conditions and reactants, the same reactant can be covered by different reaction conditions.

[0017] Optionally, after performing dimensionality reduction transformation on the reactants covered by all conditions, the process also includes reactant verification, including: determining whether the convex hull volume of the 3D point cloud formed by the 3D points of the reactants covered by all conditions can be calculated using the Python library function ConvexHull(); if not, storing all reaction conditions into a condition set and outputting the condition set; if yes, proceeding to the next step.

[0018] Optionally, if, during the iteration process, a certain reaction condition is added to the condition set but the convex hull volume of the condition set cannot be calculated using the Python library function ConvexHull(), then that condition is skipped.

[0019] Optionally, the method for calculating the coverage includes: calculating the convex hull volume and distance density of the condition set, and calculating the coverage based on the convex hull volume and distance density. The calculation expression for the coverage is as follows:

[0020] Where cover represents the coverage, ρ represents the distance density of the condition set, V represents the convex hull volume of the condition set, and ρ space V represents the distance density of reactants covered by all conditions. max This represents the convex hull volume of the reactants covered by all conditions.

[0021] Optionally, the convex hull volume is the volume occupied by a convex polyhedron formed by a set of points in three-dimensional space, wherein the set of points is a condition set, the condition set contains at least three kinds of reactants, and the molecular fingerprint of the reactants, after dimensionality reduction, contains at least three kinds of three-dimensional points. The convex hull volume is obtained by calculating the volume of the three-dimensional point cloud formed by the three-dimensional points in three-dimensional space.

[0022] Optionally, the distance density is the average distance between each point in the point cloud data and all the remaining points, wherein the point cloud data is a three-dimensional point cloud formed by the three-dimensional points corresponding to all reactants in the condition set, and the distance density is obtained by calculating the average distance between each three-dimensional point in the condition set and all the remaining points.

[0023] Optionally, the expression for calculating the convex hull volume is as follows:

[0024] Where V represents the convex hull volume of the condition set, f abc d represents the surface formed by any three points corresponding to the reactants covered by the condition set. Gf f represents the point corresponding to the reactants covered by the condition set. abc From a point outside the plane to f abc The distance between surfaces, S(f) represents f abc The area of ​​the surface.

[0025] Optionally, the distance density is calculated using the following expression:

[0026] Where ρ represents the distance density of the condition set, n represents the number of reactants covered by all conditions in the condition set, and n is an integer greater than 3, and distance(i,j) represents the distance between point i and point j.

[0027] Optionally, the step of calculating the impact of adding a single reaction condition to the condition set on the coverage of the condition set includes: inputting the reactants covered by the current reaction condition into the condition set, calculating the change in coverage before and after the condition set, and the greater the change, the greater the impact; the expression for calculating the change in coverage before and after the condition set is as follows:

[0028] Δcover = cover B -cover A ;

[0029] Where Δcover represents the change in coverage, cover A The term "cover" indicates the coverage of the condition set before the reaction conditions are added. B This indicates the coverage of the condition set after the reaction conditions are added.

[0030] Optionally, when different reaction conditions are added to the condition set and the change in coverage is the same, a reaction condition is randomly selected to be added to the condition set.

[0031] Optionally, the dimensionality reduction transformation of the reactants includes: calculating the molecular fingerprint of the reactants and reducing the molecular fingerprint to obtain three-dimensional points.

[0032] Optionally, the method for dimensionality reduction transformation of the reactants includes one of the following algorithms: PCA, LDA, ICA, SVD, LFA, TSNE, and LPP; the type of molecular fingerprint includes one of Morgan2, Morgan3, MACCS, and rxnfp.

[0033] This application also provides an electronic device, including:

[0034] Memory, used to store computer programs;

[0035] A processor is used to implement the above-described conditional filtering method when executing the computer program.

[0036] The advantages of this application compared to the prior art are:

[0037] This application provides a conditional screening method based on literature data, which can obtain reaction conditions and reactant data from massive amounts of literature data (especially high-throughput data), avoiding the tedious process of manually browsing or screening one by one, and greatly improving the efficiency of research work. Based on this, by matching reaction conditions with the reactants they cover, reducing the dimensionality of reactants to three-dimensional points and initializing the condition set and its coverage; calculating the impact of adding a single reaction condition to the condition set on the condition set coverage, selecting the reaction condition with the greatest impact to add to the condition set, updating the condition set and its coverage, iterating until the number of reaction conditions added reaches a set value or all reaction conditions have been added to the condition set, and finally outputting the condition set, it helps users obtain the combination of reaction conditions covering the most types of reactants within a specified number of literature data. This avoids the time-consuming and laborious problems of manual operation in related technologies, and successfully applies computer technology to the reaction condition screening of high-throughput chemical data, greatly promoting the integration of chemistry and computers, and contributing to the automation and intelligence of chemical development. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 is a flowchart illustrating the conditional filtering method according to an embodiment of this application;

[0040] Figure 2 is a three-dimensional spatial diagram of the conditional screening method of this application, in which the reactants are reduced in dimension to three-dimensional points. Detailed Implementation

[0041] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this application.

[0042] As shown in Figure 1, a conditional filtering method based on literature data includes:

[0043] Obtain the reaction conditions (including information such as catalysts, ligands, solvents, amounts, and temperatures used in the reaction conditions to distinguish different reaction conditions) and reactants from the literature data, traverse each reaction condition, and match the reactants covered by the current reaction condition.

[0044] Initialize the condition set and its coverage;

[0045] All reactants covered by all conditions were reduced in dimension to three-dimensional points, as shown in Figure 2;

[0046] Calculate the impact of adding a single reaction condition to the condition set on the condition set coverage, select the reaction condition with the greatest impact to add to the condition set, update the condition set and its coverage, and iterate until the number of reaction conditions added reaches the set value or all reaction conditions have been added to the condition set.

[0047] Output the condition set.

[0048] It is important to note that the reaction conditions and reactants obtained from literature data are unique. Duplicate reaction conditions and reactants will be deleted. However, when matching reaction conditions and reactants, the same reactant can be covered by different reaction conditions. In the condition set, even if a certain reactant appears in multiple reaction conditions, its coverage is only calculated once using its three-dimensional points.

[0049] Optionally, in this embodiment, the condition set is obtained through continuous iteration until the number of added reaction conditions reaches a set value or all reaction conditions have been added to the condition set. The iteration process specifically involves: calculating the impact of adding a single reaction condition to the condition set on the condition set coverage; selecting the reaction condition with the greatest impact to add to the condition set (reaction conditions already added to the condition set will not participate in the next iteration); updating the condition set and its coverage; and then proceeding to the next iteration until the number of added reaction conditions reaches the set value or all reaction conditions have been added to the condition set. In this way, a user-specified number (i.e., a set value) of reaction conditions is obtained. At this number, the coverage of the condition set (i.e., the number of reactants included) is also maximized. It is foreseeable that the condition set obtained in this way not only meets the user's requirements for the number of reaction conditions but also better suits the user's experimental needs.

[0050] In this embodiment, after dimensionality reduction transformation of all reactants covered by all conditions, reactant verification is also included, including: determining whether the convex hull volume of the 3D point cloud formed by the 3D points of all reactants covered by all conditions can be calculated by the Python library function ConvexHull(). If not, all reaction conditions are stored in the condition set and the condition set is output. If yes, proceed to the next step.

[0051] This step avoids extreme cases where the 3D points obtained from the molecular fingerprint dimensionality reduction of reactants cannot be calculated due to the inability to calculate the convex hull volume (and anomalies can be marked for the condition set in such cases as needed). For example, the number of reactants is less than or equal to 3 (when it is greater than 3, both the convex hull volume and distance density can be calculated), or the 3D points after the molecular fingerprint dimensionality reduction of reactants are on the same straight line, or the 3D points after the molecular fingerprint dimensionality reduction of reactants are on the same plane.

[0052] In this embodiment, if, during the iteration process, a certain reaction condition is added to the condition set but the convex hull volume of the condition set cannot be calculated using the Python library function ConvexHull(), then that condition is skipped.

[0053] In this step, the situation where the convex hull volume of the condition set cannot be calculated using the Python library function ConvexHull() after the reaction conditions are added to the condition set mainly occurs during the first iteration. To avoid the computer being unable to effectively iterate and output the condition set in this situation, the corresponding reaction conditions are automatically skipped in this case to ensure the smooth progress of the iteration process.

[0054] In this embodiment, the method for calculating the coverage includes: calculating the convex hull volume and distance density of the condition set, and calculating the coverage based on the convex hull volume and distance density. The calculation expression for the coverage is as follows:

[0055] Where cover represents the coverage, ρ represents the distance density of the condition set, V represents the convex hull volume of the condition set, and ρ space V represents the distance density of reactants covered by all conditions. max This represents the convex hull volume of the reactants covered by all conditions.

[0056] The convex hull volume refers to the volume occupied by a convex polyhedron formed by a set of points in three-dimensional space. In this embodiment, the set of points is the condition set, which contains at least three kinds of reactants. After the molecular fingerprint of the reactants is reduced in dimension, it contains at least three kinds of three-dimensional points. The convex hull volume is obtained by calculating the volume of the three-dimensional point cloud (convex hull) formed by the three-dimensional points in three-dimensional space.

[0057] Distance density is typically used to describe the average distance between each point in point cloud data and all other points. In this embodiment, the point cloud data refers to the three-dimensional point cloud formed by the three-dimensional points corresponding to all reactants in the condition set. The distance density is obtained by calculating the average distance between each three-dimensional point in the condition set and all other points.

[0058] In this embodiment, the expression for calculating the convex hull volume is as follows:

[0059] Where V represents the convex hull volume of the condition set, f abc d represents the surface formed by any three points corresponding to the reactants covered by the condition set. Gf f represents the point corresponding to the reactants covered by the condition set. abc From a point outside the plane to f abc The distance between surfaces, S(f) represents f abc The area of ​​the surface.

[0060] In this embodiment, the calculation expression for the distance density is as follows:

[0061] Where ρ represents the distance density of the condition set, n represents the number of reactants covered by all conditions in the condition set, and n is an integer greater than 3, and distance(i,j) represents the distance between point i and point j.

[0062] In this embodiment, the step of calculating the impact of adding a single reaction condition to the condition set on the coverage of the condition set includes: inputting the reactants covered by the current reaction condition into the condition set, calculating the change in coverage before and after the condition set, and the greater the change, the greater the impact; the expression for calculating the change in coverage before and after the condition set is as follows:

[0063] Δcover = cover B -coverA ;

[0064] Where Δcover represents the change in coverage, cover A The term "cover" indicates the coverage of the condition set before the reaction conditions are added. B This indicates the coverage of the condition set after the reaction conditions are added.

[0065] It should be noted that when different reaction conditions are added to the condition set and the coverage changes by the same amount, one reaction condition is randomly selected to be added to the condition set.

[0066] In this embodiment, the dimensionality reduction transformation of the reactants includes: calculating the molecular fingerprint of the reactants and reducing the molecular fingerprint to obtain three-dimensional points.

[0067] By calculating the molecular fingerprint of the reactants and reducing the molecular fingerprint to three-dimensional points, reaction conditions can be screened in subsequent steps such as calculating coverage.

[0068] In this embodiment, the method for dimensionality reduction transformation of the reactants includes one of the following algorithms: PCA, LDA, ICA, SVD, LFA, TSNE, and LPP (PCA is selected in this embodiment); the molecular fingerprint type includes one of Morgan2, Morgan3, MACCS, and rxnfp.

[0069] This application also provides an electronic device, including:

[0070] Memory, used to store computer programs;

[0071] A processor is used to implement the above-described conditional filtering method when executing the computer program.

[0072] The present application has been further described above with reference to specific embodiments. However, it should be understood that the specific descriptions herein should not be construed as limiting the substance and scope of the present application. Various modifications made by those skilled in the art to the above embodiments after reading this specification are all within the scope of protection of the present application. Industrial applicability

[0073] This application provides a condition screening method and electronic device based on literature data, which can obtain reaction conditions and reactant data from massive literature data, avoiding the tedious process of manually browsing or screening one by one, and greatly improving the efficiency of research work.

[0074] Furthermore, it is understood that the condition screening method and electronic equipment based on literature data in this application are reproducible and can be widely applied in the field of chemical reaction condition screening technology.

Claims

1. A conditional screening method based on literature data, characterized in that, include: Obtain the reaction conditions and reactants from the literature data, iterate through each reaction condition, and match the reactants covered by the current reaction condition. Initialize the condition set and its coverage; All reactants covered by all conditions are reduced in dimension and converted into three-dimensional points; Calculate the impact of adding a single reaction condition to the condition set on the condition set coverage, select the reaction condition with the greatest impact to add to the condition set, update the condition set and its coverage, and iterate until the number of reaction conditions added reaches the set value or all reaction conditions have been added to the condition set. Output the condition set.

2. The conditional screening method according to claim 1, characterized in that, After dimensionality reduction transformation of all reactants covered by all conditions, reactant verification is also included, including: determining whether the convex hull volume of the 3D point cloud formed by the 3D points of all reactants covered by all conditions can be calculated by the Python library function ConvexHull(). If not, all reaction conditions are stored in the condition set and the condition set is output. If yes, proceed to the next step.

3. The conditional screening method according to claim 1, characterized in that, If, during the iteration process, a certain reaction condition is added to the condition set but the convex hull volume of the condition set cannot be calculated using the Python library function ConvexHull(), then that condition is skipped.

4. The conditional screening method according to claim 1, characterized in that, The method for calculating the coverage includes: calculating the convex hull volume and distance density of the condition set, and calculating the coverage based on the convex hull volume and distance density. The calculation expression for the coverage is as follows: Where cover represents the coverage, ρ represents the distance density of the condition set, V represents the convex hull volume of the condition set, and ρ space V represents the distance density of reactants covered by all conditions. max This represents the convex hull volume of the reactants covered by all conditions.

5. The conditional screening method according to claim 4, characterized in that, The expression for calculating the convex hull volume is as follows: Where V represents the convex hull volume of the condition set, f abc d represents the surface formed by any three points corresponding to the reactants covered by the condition set. Gf f represents the point corresponding to the reactants covered by the condition set. abc From a point outside the plane to f abc Distance between surfaces S(f) represents f abc The area of ​​the surface.

6. The conditional filtering method according to claim 4, characterized in that, The expression for calculating the distance density is as follows: Where ρ represents the distance density of the condition set, n represents the number of reactants covered by all conditions in the condition set, and n is an integer greater than 3, and dis tan ce(i,j) represents the distance between point i and point j.

7. The conditional screening method according to claim 1, characterized in that, The step of calculating the impact of adding a single reaction condition to the condition set on the coverage of the condition set includes: inputting the reactants covered by the current reaction condition into the condition set, calculating the change in coverage before and after the condition set is calculated; the greater the change, the greater the impact. The expression for calculating the change in coverage before and after the condition set is as follows: Δcoverer = coverer B -cov er A ; Where Δcov er represents the change in coverage, cov er A This indicates the coverage of the condition set before the reaction conditions were added. B This indicates the coverage of the condition set after the reaction conditions are added.

8. The conditional screening method according to claim 1, characterized in that, The dimensionality reduction transformation of the reactants includes: calculating the molecular fingerprint of the reactants and reducing the molecular fingerprint to obtain three-dimensional points.

9. The conditional screening method according to claim 8, characterized in that, The method for dimensionality reduction transformation of the reactants includes one of the following algorithms: PCA, LDA, ICA, SVD, LFA, TSNE, and LPP; the type of molecular fingerprint includes one of Morgan2, Morgan3, MACCS, and rxnfp.

10. The conditional screening method according to claim 1, characterized in that, The reaction conditions include information such as the catalyst, ligand, solvent, dosage, and temperature used in the reaction.

11. The conditional filtering method according to claim 1, characterized in that, The reaction conditions and reactants obtained from literature data are unique. Duplicate reaction conditions and reactants will be deleted. However, when matching reaction conditions and reactants, the same reactant can be covered by different reaction conditions.

12. The conditional screening method according to claim 4, characterized in that, The convex hull volume is the volume occupied by a convex polyhedron formed by a set of points in three-dimensional space. The set of points is a condition set, which contains at least three reactants. The molecular fingerprint of the reactants, after dimensionality reduction, contains at least three three-dimensional points. The convex hull volume is obtained by calculating the volume of the three-dimensional point cloud formed by the three-dimensional points in three-dimensional space.

13. The conditional screening method according to claim 4, characterized in that, The distance density is the average distance between each point in the point cloud data and all the remaining points. The point cloud data is a three-dimensional point cloud formed by the three-dimensional points corresponding to all reactants in the condition set. The distance density is obtained by calculating the average distance between each three-dimensional point in the condition set and all the remaining points.

14. The conditional screening method according to claim 7, characterized in that, When different reaction conditions are added to the condition set and the coverage changes by the same amount, a reaction condition is randomly selected and added to the condition set.

15. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the conditional filtering method as described in any one of claims 1 to 14.

Images