CN112365989B

CN112365989B - Equivalent signal mining method for SRS combined adverse reaction signals

Info

Publication number: CN112365989B
Application number: CN202011304774.5A
Authority: CN
Inventors: 袁洪; 刘心瑶; 吴俏玉; 陆瑶; 蔡菁菁; 李莹; 黄志军
Original assignee: Changsha Hongyuan Cardiovascular Health Research Institute
Current assignee: Changsha Hongyuan Cardiovascular Health Research Institute
Priority date: 2020-11-19
Filing date: 2020-11-19
Publication date: 2022-05-03
Anticipated expiration: 2040-11-19
Also published as: CN112365989A

Abstract

The invention discloses an equivalent signal mining method facing an SRS combined adverse reaction signal, which comprises the steps of acquiring data through an SRS spontaneous presentation system of adverse drug events to obtain a combined adverse reaction signal set X ═ D₁,D₂,…,D_x}，D_xA drug combination which is judged to be an adverse reaction signal; preferentially processing the signal with the minimum size of the drug combination set and solving the maximum equivalence group D of the D^GCombining with D^GThere is a maximum equivalence set to a maximum equivalence set D of the intersection^G(ii) a Will no longer belong to D^GAnd a signal equivalent to D is incorporated into D^GD is^GAdding X^GOutputting the maximum equivalence set X^G。X^GEach maximum equivalence group in the group is a signal set with pairwise equivalence relation between the combined adverse reaction signals, and one signal is selected from the maximum equivalence groups to represent the signal set (namely the maximum equivalence group) to participate in other data analysis, so that the workload of the overall analysis is reduced.

Description

Equivalent signal mining method for SRS combined adverse reaction signals

Technical Field

The invention relates to the technical field of signal mining, in particular to an equivalent signal mining method for adverse reaction signals of SRS combined medication, wherein SRS represents a spontaneous report system for adverse drug events.

Background

The study on the interaction of all the combined drugs cannot be completed in the pre-marketing stage due to the factors such as the experimental range, the study object, the time and the like, and many potential drug interactions can be discovered only in the process of large-scale and long-time use after the drugs are marketed. The Spontaneous drug adverse event Reporting System (SRS) provides an important data source for drug interaction signal mining after marketing. For example, the FDA Adverse Event Reporting System (FAERS) in the united states reflects the complexity of medication safety in real life, and it is investigated that 60% of Adverse Event reports have more than one drug, 70% have more than one drug, and 84% have at least 3 total drugs or Adverse events. These features present opportunities and challenges for combination drug adverse reaction signal mining.

In order to break through the dilemma of 'mass data and lack of information', mainstream researches utilize a computer to discover adverse reaction signals of combined medication in batches in spontaneous presentation system data. The signal contains two elements: drug combination D and target adverse reaction AE. A drug combination is a collection of drugs, which may also be referred to as a combination or co-drug. If combination D is a signal for an adverse reaction AE, it means that the targeted adverse reaction AE may occur when the patient is taking all of the drugs in D at the same time. When AE is determined, the signal can be expressed as a drug combination. It should be noted that the signal is only a clue and does not prove the causal relationship between the drug combination and the adverse reaction, which needs more complete medical experiment and mechanism analysis to confirm that the signal that can be confirmed is called a positive signal and the signal that cannot be confirmed is called a false positive signal. The higher the positive signal percentage in the signal, the higher the quality of signal mining. The signal strength metric is a key to determining the quality of signal mining. The imbalance measurement is the basic idea of signal metrics, i.e., "imbalance" or "dissimilarity" of an event of interest as compared to other events. The method comprises two major categories, namely a frequency method and a Bayesian method, wherein the frequency method comprises a relative hazard ratio (RR), a ratio report ratio (PRR), a report ratio (ROR) and the like, and the Bayesian method comprises a Bayesian confidence degree progressive neural network (BCPNN), an empirical Bayesian gamma Poisson distribution reduction Method (MGPS), a BCPNN high-dimensional expansion version, an omega contraction measurement method and the like. The methods have advantages, but on the whole, the common problems of large signal clue quantity and low accuracy (4%) exist, the advantages of large data cannot be truly exerted, and the defect is more prominent when high-order combined drug adverse reaction signals are mined. This is because the real-life drug combination mode is very complicated, and the known imbalance measurement method estimates the risk of adverse signal reaction by assuming independent and irrelevant drug combination, only evaluates the drug combination without considering the influence of other drug combinations, so that other drugs (combination) daily combined with positive signals can be estimated as false positive signals. One of the very typical problems is that several apparently distinct combination adverse reaction signals are expressed from the same set of patient reports in the SRS database. The presence of such signals can cause two problems for subsequent signal analysis by medical researchers: (1) performing data analysis independently on signals from the same patient subgroup can result in duplication of labor; (2) from the point of view of data analysis, these signals are confounding each other, in other words, the number of positive signals therein is much smaller than the number of signals. At present, how to describe and mine such signals in a normative way, no relevant research work can be used for reference.

Disclosure of Invention

Technical problem to be solved

Aiming at the problems, the invention provides an equivalent signal mining method facing an adverse reaction signal of SRS combined medication. If the coincidence degree of the patient groups corresponding to the two signals is high, case difference is insufficient to support signal intensity judgment, the two signals are equivalent to the equivalent description of adverse drug reactions of the same patient group, and the two signals are called to have an equivalent relationship, so that the invention provides a judgment method. On the basis, a transmission hypothesis and a combination hypothesis are adopted, and a signal set with pairwise equivalence relation between the combined adverse reaction signals is excavated, and the signal set is called as a maximum equivalence group. And selecting a signal from the maximum equivalence group to represent the whole set to participate in other data analysis, so that the workload of the whole analysis is reduced, and the equivalent signal of the signal can be regarded as a confounding factor, which is beneficial to improving the quality of the signal analysis.

(II) technical scheme

Based on the technical problems, the invention provides an equivalent signal mining method for adverse reaction signals of SRS combined medication, which comprises the following steps:

s1, making the unprocessed set Z as X, and making the maximum equivalence set X of the combined drug adverse reaction signal set X^GIn the empty set, the combined adverse reaction signal set X acquires data and preprocesses the data through an SRS (sounding reference signal) of a drug adverse event spontaneous presentation system, and then the combined adverse reaction signal set X ═ { D ═ is obtained by mining₁,D₂,…,D_x}，D_xA drug combination for determining an adverse reaction signal;

s2, judging whether Z is an empty set, if so, entering a step S9, and otherwise, entering a step S3;

s3, priority processing the signal with the smallest medicine combination size, wherein the medicine combination size represents the number of medicines contained in the medicine combination, i.e. processing the signal

If | represents the size of the set, the unprocessed set Z ═ Z- { D }, and D is initialized^G＝{D}；

S4, judgment X^GIf the current set is an empty set, if so, the step S6 is executed, otherwise, the step S5 is executed;

s5, traverse X^GCombining with D^GThere is a maximum equivalence set to a maximum equivalence set D of the intersection^G；

S6, judging whether Z is an empty set, if so, entering a step S8, and if not, entering a step S7;

s7, traversing Z, adding the signals equivalent to D into the maximum equivalent group D of D^G；

S8, order X^G＝X^G∪{D^GReverting to step S2;

s9, outputting the maximum equivalence group set X^GThat is, the adverse side effect of combined medicationThe method comprises the steps of collecting maximum equivalence groups of response signals, wherein each maximum equivalence group is a combined adverse reaction signal collection with pairwise equivalence relation, selecting a signal, namely a drug combination, from each maximum equivalence group, and representing the maximum equivalence group to participate in further analysis of combined adverse reaction data, wherein the data are obtained from demographic information, medication information and adverse reaction information of a spontaneous drug adverse event report system SRS.

Further, the method for determining the drug combination with adverse reaction signals comprises the following steps: if the frequency of the target adverse reaction signals AE reported by the patients taking the medicine combination D is greater than the support threshold; the length of the drug combination D is not more than the drug combination signal length threshold MAX _ D; d obtained from SRS data of PS patient set is used as lower bound Q (PS, D, AE) CI of signal intensity confidence interval of target adverse reaction signal AE^->A signal strength threshold δ; d is determined to be the drug combination of the targeted adverse reaction signal AE.

Further, Q (PS, D, AE) is a signal metric function of the non-equilibrium metric class, and the metric methods adopted include RR, PRR, RoR, BCPNN, and MGPS.

Further, the preprocessing comprises data deduplication and drug name normalization.

Further, the step S5 includes the following steps:

s5.1, traversing X^GSelect not traversed member D_k ^GJudgment of D_k ^GAnd D^GIf not, go to step S5.2, if yes, D^G＝D^G∪D_k ^G，X^G＝X^G-{D_k ^GStep S5.2;

s5.2, judging and traversing X^GIf not, the process goes to step S6, otherwise, the process goes to step S5.1.

Further, the step S7 includes the following steps:

s7.1, selecting non-traversed member D of Z_i；

S7.2, judging whether D is equivalent to D_iIf yes, let D^G＝D^G∪{D_iStep 7.3, if not, step 7.3 is carried out;

s7.3, judging whether the traversal Z is finished, if not, entering the step S7.1, and if so, indicating D^GAfter the solution is completed, the process proceeds to step S8.

Further, it is determined in step S7.2 whether D is equivalent to D_iThe method comprises the following steps:

s7.2.1, and Q (PS, D, AE) CI^-≤Q(PS,D_i,AE).CI⁺And Q (PS, D)_i,AE).CI^-≤Q(PS,D,AE).CI⁺If yes, go to step S7.2.2; otherwise, go to step S7.3; wherein Q (e.,. e.) represents a signal metric function, Q (PS, D, AE)⁺，Q(PS,D,AE).CI^-Representing the upper and lower bounds of a signal intensity confidence interval of a target adverse reaction signal AE, which is D obtained according to SRS data of a patient set PS;

s7.2.2, determination of absolute values | C (G (D), AE) -C (G (D)_i) AE) | is greater than θ, if yes, go to step S7.3, otherwise go to step S7.2.3; wherein G (·) represents a medication case function, C (,) represents an adverse reaction report statistical function, C (G (D), AE) represents the number of reports of adverse reactions AE from a case set G (D) taking a drug combination D, and θ represents an adverse reaction report number threshold;

s7.2.3, determining whether D is D_iIf so, go to step S7.2.4, otherwise go to step S7.2.5;

s7.2.4, let C (G (D) -G (D)_i),AE)＝C(G(D),AE)-C(G(D_i),AE)，C(G(D_i) -g (d), AE) ═ 0, go to step S7.2.6;

s7.2.5, determining whether | D | is equal to the medication combination signal length threshold MAX _ D, if yes, entering step S7.2.6; otherwise, go to step S7.3;

s7.2.6, traversing the SRS data, and counting to obtain C (G), (D) -G (D)_i) AE) and C (G (D)_i) -values of g (d), AE);

s7.2.7, judging C (G), (D) -G (D)_i)，AE)<θ and C (G (D)_i)-G(D)，AE)<Whether or not θ becomesIf both are true, go to step S7.2.8; otherwise, go to step S7.3;

s7.2.8, D is equivalent to D_iLet D^G＝D^G∪{D_i}。

The invention discloses a server, comprising:

at least one processor; and at least one memory communicatively coupled to the processor, wherein:

the memory stores program instructions executable by the processor, and the processor invokes the program instructions to perform the equivalent signal mining method for SRS-oriented combined adverse drug reaction signals.

The invention discloses a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the equivalent signal mining method for an adverse reaction signal of an SRS combination.

(III) advantageous effects

The technical scheme of the invention has the following advantages:

(1) the invention first takes the value D from X and solves the maximum equivalence group D of D^GAdding X^GThen, the value D is re-taken and X is traversed^GMerging the maximum equivalent sets with intersection, and solving the new maximum equivalent set D of D^GIncorporating X^GRepeating the steps in the cycle until each drug combination in the X has been valued, and obtaining the final maximum equivalent group set X^G；X^GEach maximum equivalence group in the drug combination adverse reaction signal set is a drug combination adverse reaction signal set with pairwise equivalence relation, and one signal is selected from the set according to an equivalence substitution assumption so as to represent the whole set to participate in other data analysis, thereby being beneficial to reducing the workload of signal analysis; when finer-grained data analysis is carried out on a specific signal, an equivalent signal of the signal can be used as a confounding factor, and the signal analysis quality is improved;

(2) step S7 of the present invention is to solve the maximum equivalence group D of D according to the equivalence relation determination method^GIn the process, traverseZ, adding the signal equivalent to D to the maximum equivalence group D of D^GThe accuracy is high, and before the equivalence relation judging method is utilized, special conditions that equivalence relation judgment is not established without inquiring SRS data are eliminated, SRS data inquiry is avoided, and the calculation efficiency is improved;

(3) in step S7, the equivalence relation is determined by using the method of | D | and the size of the combined medication signal length threshold MAX _ D in combination with the equivalence transfer hypothesis and the equivalence combination hypothesis, so that the execution of equivalence relation determination is avoided, SRS data query is avoided, and the calculation efficiency is further improved;

(4) step S5 of the present invention at X^GAfter non-empty set, traverse X^GMerging and D according to the equal-cost transfer assumption^GThere is a maximum equivalence set to a maximum equivalence set D of the intersection^GSo that and D^GSignals contained in the maximum equivalent group with intersection do not need to be subjected to equivalent judgment one by one, so that the number of relation judgment times in the equivalent group is reduced, the calculation amount is reduced, and the analysis efficiency is improved.

Drawings

The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:

FIG. 1 is a schematic overall flow chart of a combined medication adverse reaction equivalence relation signal mining method based on SRS data according to an embodiment of the invention;

FIG. 2 is an example of equivalence relations and maximum equivalence sets for an embodiment of the present invention;

FIG. 3 is a flowchart of a signal mining method for adverse reaction equivalence relation based on SRS combined medication according to an embodiment of the invention;

FIG. 4 is a flowchart of an embodiment of the present invention for determining whether D is equivalent to D_iA flow chart of the method of (1).

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

The invention relates to a mining method for analyzing adverse reaction equivalence relation signals of combined drug bleeding by using an adverse event report file issued by FAERS, which comprises the following steps of:

in the stage I, data are acquired and preprocessed through an SRS (sounding reference signal) of a drug adverse event spontaneous reporting system: data are acquired through an SRS (sounding reference signal) of a drug adverse event spontaneous presentation system, data are deduplicated, drug names are normalized, and target adverse reactions AE are selected. This is the basis and precondition for all signal mining efforts. After data preprocessing, the signal is analyzed in the last two stages. For the purposes of the following expressions, the following symbols are defined: let the total patient set in the SRS be PS ═ p₁,p₂,…,p_mAll drugs are set as DS ═ d }₁,d₂,…,d_n}. Let C (,) denote the adverse reaction reporting statistical function, given a subset of patients P, C (P, AE) denotes the number of reports of adverse reaction AEs from subset of patients P. Let G (.) denote the medication case function, given signals D, G (D) as a set of cases taking medication combination D.

Obtaining adverse event report files published by FAERS 2018 three-quarter data to obtain 420,915 case reports, 1,651,966 medication records (including 51083 medicine names) and 1,329,530 adverse event records (including 11,944 adverse event records), integrating the demographic information, medication information and adverse drug reactions related by case IDs, and carrying out standardized processing on the adverse drug reaction names and the medicine names. Bleeding events were selected as target adverse events AE, involving 194 adverse events (all generalizable to bleeding events but stated differently) for a total of 19,067 bleeding event records.

And II, excavating adverse reaction signals of the combined medicine: for selected SRS data and target adverse reaction AE, excavating combined drug adverse reaction signals to obtain a combined drug adverse reaction signal set X ═ D₁,D₂,…,D_xIn which the signal D is_iIs a combination of drugs (drug set) taken simultaneously by patients_iFor each drug in (1), the patient is said to take D_i. . Wherein the signal evaluation is jointly determined by the following three parameters: (1) threshold of supportSupport: the frequency of target adverse reaction AE reported by the medication patients from the drug combination D is more than Support; (2) length threshold MAX _ D: the length of drug combination D (i.e., the number of drugs contained in D) is not greater than MAX _ D, i.e., | D<MAX _ D; (3) signal metric function Q (,) and signal strength threshold δ: the function Q is input into a patient set PS, a drug combination D and a target adverse reaction AE, and Q (P, D, AE) is output as a Confidence Interval (CI) of signal intensity of an AE signal of D obtained from SRS data of the patient set PS, wherein CI is [ CI ]^-，CI⁺]If the confidence interval is lower bound CI^->And delta, D is judged as a signal of the target adverse reaction AE. The signal measurement function Q can adopt a non-equilibrium measurement method, which is a category of traditional research work, and the known measurement methods include a relative hazard ratio (RR), a ratio report ratio (PRR), a report ratio (ROR), a bayesian confidence degree progressive neural network (BCPNN), an empirical bayesian gamma poisson distribution subtraction Method (MGPS), and the like, and the corresponding relationship between Q and a typical method is shown in table 1. When solving Q (PS, D, AE) by the non-equilibrium metric method, C (g, (D), AE) needs to be calculated based on SRS data. In order to increase the efficiency of the subsequent calculation, each signal D in the set X has additional information Q (PS, D, AE) and C (g (D), AE).

TABLE 1 exemplary Signal metric method and confidence interval selection therefor

Drug combination bleeding event signals were mined using the FAERS 2018 three-quarter data. Let the Support threshold Support be 20, the length threshold MAX _ D be 3, the signal metric function Q (·) be RoR, and the signal strength threshold δ be 2, to obtain 1,530 drug combination bleeding event signals.

And III, excavating an adverse reaction equivalent relation signal of the combined medication from the adverse reaction signal set of the combined medication: the input of the stage is a signal set with pairwise equivalence relation between the combined drug adverse reaction signals to be mined, wherein the pairwise equivalence relation means that the patient groups corresponding to the two signals have high coincidence degree, case difference is insufficient to support signal intensity judgment, and the two signals are equivalent to equivalent description of the same patient group. The signal set obtained in the second stage is analyzed, the equivalence relation among the signals is identified, and the maximum equivalent signal set is obtained. This stage generally employs the same signal metric function Q as the second stage.

In order to disclose the characteristics of the equivalence relation, the ideal equivalence relation is analyzed firstly, the inference of the ideal equivalence relation is extended to the hypothesis under the general condition, and then the equivalence relation judgment standard and the mining algorithm under the general condition are obtained:

the ideal equivalent relationship is as follows: given signal D_a，D_bIf G (D)_a)＝G(D_b) G (-) denotes the case function, drug combination D_aAnd D_bThe same set of cases is called D_aAnd D_bHave an equivalence relation between them, is called D_aIs equivalent to D_b. From the definition of the ideal equivalence relation, the following reasoning holds:

inference 1 equivalent substitutions: if signal D_aIs equivalent to D_bFor a given signal metric function Q, patient set P and signal D_cThen the following equation holds: (i) q (P, D)_a,AE)＝Q(P,D_bAE), namely under the same conditions, D_aAnd D_bThe signal intensities are consistent; (ii) q (P-G (D)_a),D_c,AE)＝Q(P-G(D_b),D_cAE), namely under the same conditions, D_a、D_bThe effect on the other signal evaluations was consistent.

Equivalence substitution inference indicates that for an equivalent signal D_a，D_bOnly one of the representatives needs to be selected to participate in the signal correlation analysis, which reduces the workload of the overall analysis.

Inference 2 equivalent transfer: if signal D_aEquivalent to signal D_bAnd D is_bEquivalent to signal D_cThen D is_aIs equivalent to D_c；

Inference 3 merge equivalencies: if signal D_aEquivalent to signal D_bOrder signal

And is

Then D is_cIs equivalent to D_a。

However, the ideal equivalence relation is a special case, and the possibility of the ideal equivalence relation among signals is very little in view of the complexity of the combined medication in reality. For the assessment of adverse drug reaction signals, a small number of adverse reaction report differences were not statistically significant. Therefore, the equivalence between the determination signals focuses on two points: (1) the signal strengths have no significant difference, namely, the confidence intervals of the signal strengths have intersection; (2) for the target adverse reaction AE, a proper amount of difference exists between the adverse reaction reports of the allowable signals, and the difference is restrained by adopting a parameter 'adverse reaction report number threshold value theta'. The equivalence relation decision is defined as follows.

And (3) judging equivalence relation: for a given signal D_a，D_bIf the following 2 conditions are satisfied, D_aAnd D_bHave an equivalence relation between them, called D_aIs equivalent to D_b. : (i) there was no significant difference between the two signal strength estimates, i.e., Q (PS, D)_a,AE).CI^-≤Q(PS,D_b,AE).CI⁺，Q(PS,D_b,AE).CI^-≤Q(PS,D_a,AE).CI⁺. (ii) Adverse case record differences between the two signals were less than θ, i.e., C (G (D)_a)-G(D_b)，AE)<＝θ，C(G(D_b)-G(D_a)，AE)<θ. Wherein the set subtraction operation relationship is defined as G (D)_a)＝G(D_a)-G(D_a)∩G(D_b)。

Analyzing the above judgment process, in the condition (i), Q (PS, D)_a,AE)、Q(PS,D_bAE) is the signal D_a、D_bThe attached information of (2) is directly compared with the numerical value, so that the calculation cost is not large; in the condition (ii), in general, C (G (D)_a)-G(D_b),AE)、C(G(D_b)-G(D_a) AE) needs to query SRS data, which is high in calculation overhead. However, there are two special conditions bySignal D_a、D_bAuxiliary information C (G (D))_a),AE)、C(G(D_b) AE), SRS data may not need to be queried: (ii-1) if

Then C (G (D)_a)-G(D_b),AE)＝C(G(D_a),AE)-C(G(D_b),AE)，C(G(D_b)-G(D_a) AE) 0; (ii-2) apparently, | C (G (D)_a),AE)-C(G(D_b),AE)|<＝Max(C(G(D_a)-G(D_b),AE),C(G(D_b)-G(D_a) AE)), therefore, if | C (G (D))_a),AE)-C(G(D_b),AE)|>θ, it is understood that the condition (ii) does not hold.

Considering that the equivalence relation judgment process relates to the SRS data query, in order to further improve the calculation efficiency, the equivalence transmission inference and the merging equivalence inference of the ideal equivalence relation are used for reference, the two inferences are used as the equivalence transmission assumption and the equivalence merging assumption, and the SRS data query is avoided as much as possible. When the signal D is known from the equivalent transfer assumption_aEquivalent to signal D_bAnd D is_bEquivalent to signal D_cIn this case, D can be determined without executing the equivalence relation determination condition_aAnd D_cAnd equivalently, SRS data query is avoided. Given signal D, as can be seen from the combined equivalence assumptions_aAnd D_bWherein | D_a|<MAX _ D, if D_aIs equivalent to D_bThen there must be a signal D in the set of signals X_cSatisfy the following requirements

And D_cIs equivalent to D_aAnd D_b. In this case, pair D is discarded based on the equivalent transfer assumption_aAnd D_bD is finally obtained by judging the equivalence relation_aAnd D_bAnd (4) an equivalent conclusion. At the same time, this is also advantageous for reducing the computational overhead when the conditions are met

Then, D is judged_bAnd D_cAn equivalence relation ofSRS data queries need not be performed.

For an input signal set X, the patent intends to group the signals therein, and for selected signals, if there is pairwise equivalence between the signals, put the signals into a signal set, which is called a maximum equivalence set. The definition is as follows:

maximum equivalence set definition: given set of signals

If for D^GOf any two signals D_a，D_bSatisfy D_aEquivalent D_bSet of signals D^GReferred to as an equivalence group. If there is no equivalence group D^G’So that

Then D is^GReferred to as the maximum equivalence group.

Based on the equivalence transfer assumption, it can be demonstrated that any one signal belongs to at most one maximum equivalence group. As shown in fig. 2, the inference of the ideal equivalence relation is extended to the hypothesis under the general situation, and then the equivalence relation decision criterion and the mining algorithm under the general situation are obtained: shown, { D, { when no equivalent transfer assumptions are assumed₁,D₂,D₃And { D }₁,D₂,D₄Are the largest equivalence sets. Wherein D is₁，D₂Belong to the two largest equivalence groups, thus creating a conceptual divergence, according to the equivalence transfer assumption, D₁，D₂Belong to a maximum equivalence group { D₁,D₂,D₃,D₄}。

In summary, the signal set X is divided into several maximal equivalence groups without intersection, and the set of these maximal equivalence groups is called the maximal equivalence group set X^G。

When SRS data and a signal set X (including Q (PS, D, AE) and C (G (D) related to each signal) are inputted_a) AE) data) as follows, as shown in fig. 3:

s1, making Z as X in the untreated set, and making X as X in the combined adverse reaction signal setMaximum equivalence set X^GIs an empty set phi;

s2, judging whether Z is an empty set, if so, entering a step S9, and if not, entering a step S3;

if Z is an empty set, the combined drug adverse reaction signal set X is completely processed;

s3, priority processing the signal with the smallest size of the medicine combination (the number of medicines contained in the medicine combination), i.e. order to process the signal

Initialization D^GWhere | represents the size of the set (i.e., the number of elements in the set), then the unprocessed set Z ═ Z- { D };

if X^GIf it is an empty set, the process proceeds to step S6;

according to the equivalence transfer assumption, if two maximum equivalence groups D^GAnd D_k ^GIf there is an intersection, the signals of the two maximum equivalence sets are equivalent, so that the signals are combined into a maximum equivalence set D^GNo longer need to match D^GD with intersection_k ^GThe medicine combination is subjected to equivalence judgment one by one, the operation amount is greatly reduced through equivalence transmission hypothesis, and the analysis efficiency is improved.

S5.2, judging and traversing X^GIf yes, the step S6 is executed, otherwise, the step S5.1 is executed;

s7, traversing Z, adding the signals equivalent to D into the maximum equivalent group D of D^G(ii) a S7.1, selecting non-traversed member D of Z_i；

Judging whether D is equivalent to D in step S7.2_iAs shown in fig. 4:

s7.2.1, determining whether the following two inequalities, Q (PS, D, AE) CI, are true^-≤Q(PS,D_i,AE).CI⁺，Q(PS,D_i,AE).CI^-≤Q(PS,D,AE).CI⁺If both are true, go to step S7.2.2; otherwise, go to step S7.3;

s7.2.2, determination of absolute values | C (G (D), AE) -C (G (D)_i) AE) | is greater than θ, if yes, go to step S7.3, otherwise go to step S7.2.3;

s7.2.4, let C (G (D) -G (D)_i),AE)＝C(G(D),AE)-C(G(D_i),AE)，C(G(D_i) -g (d), AE) ═ 0, go to step S7.2.7;

s7.2.7, judging whether the following two inequalities hold: c (G (D) -G (D)_i)，AE)<＝θ，C(G(D_i)-G(D)，AE)<θ; if both are true, go to step S7.2.8; otherwise, go to step S7.3;

s7.2.8, D is equivalent to D_iLet D^G＝D^G∪{D_i}；

Step S7.2.1 is to make a decision first based on the equivalence relation decision (i), step S7.2.2 eliminates the special case condition that the equivalence relation decision (ii) that does not require querying for SRS data does not hold, and SRS data querying is avoided, and steps S7.2.3 to S7.2.6 show that when D is D_iIs equal to the combination signal length threshold MAX _ D, and if one is true, the method proceeds to step S7.2.6, wherein steps S7.2.3-S7.2.4 determine (ii-1) based on the equivalence relation, steps S7.2.5-S7.2.6 avoid the SRS data query based on the equivalence propagation assumption and the equivalence combination assumption, and step S7.2.7 further determines (ii-2) based on the equivalence relation_iEquivalent to D, the signal equivalent to D is added to the maximum equivalence set D of D in step S7.2.8^G。

S8, order X^G＝X^G∪{D^GReverting to step S2;

s9, outputting the maximum equivalence group set X^G。

The invention first takes the value D from X and solves the maximum equivalence group D of D^GAdding X^GThen, the value D is re-taken and X is traversed^GMerging the maximal equivalent sets with intersection to solve the new maximal equivalent set D of D^GIncorporating X^GRepeating the steps in the cycle until each drug combination in the X has been valued, and obtaining the final maximum equivalent group set X^G。

For 1,530 signals obtained in the previous stage, the adverse reaction report number threshold θ is set to 4, and a total of 783 maximum equivalence sets are obtained through mining. Wherein, the maximum equivalence group containing 2 and more than 2 signals has 47, and the 47 maximum equivalence groups contain 794 signals in total, when each maximum equivalence group selects a representative signal for analysis, 794 signal analysis processes can be compressed into 47. From the perspective of a single maximum equivalence group, the maximum equivalence group containing the most signals contains 399 signals,one of the signals is { hydrocinnamatoxin, risperidone }, the RoR thereof₀₂₅10.51, significantly above the signal decision threshold (δ 2). The signal has an equivalent relationship with 398 signals { levo-promazine, hydrocinnamatoxin }, { glutelin, levo-promazine }, etc., and the patient data of the 399 signals are nearly consistent, which also means that only a small number of positive signals exist. The pharmaceutical researchers need to take full consideration in experimental confirmation and mechanistic analysis of these signals, i.e. if one of the signals is selected for analysis, the other 398 signals are suggested as confounding factors.

The method can be used for analyzing the combined adverse reaction signals independently, and can also be used in cooperation with a strong signal screening method facing the SRS combined adverse reaction signals and/or an in-doubt signal mining method facing the SRS combined adverse reaction signals. The three methods are all used for analyzing the relation between signals, but the targets are respectively emphasized, strong signal screening is used for finding out a positive signal clue, doubtful signal mining is used for finding out a false positive signal clue, equivalence relation signal mining is used for finding out a signal clue with consistent data height, and the clues are respectively beneficial to improving the signal analysis quality. And (4) carrying out equivalent relation signal mining firstly, then carrying out strong signal screening and doubt signal mining execution. The equivalence relation signal mining method obtains a maximum equivalence set, only one signal in each maximum equivalence set needs to be selected to participate in subsequent analysis, and the calculation cost of the whole analysis process can be reduced.

Finally, it should be noted that the above-described method can be converted into software program instructions, and can be implemented by using a control system including a processor and a memory, or by using computer instructions stored in a non-transitory computer-readable storage medium. The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

In conclusion, the equivalent signal mining method for the adverse reaction signal of the SRS combination drug has the following advantages:

(2) step S7 of the present invention is to solve the maximum equivalence group D of D according to the equivalence relation determination method^GIn the process, Z is traversed, and the signals in which D is equivalent are added to the maximum equivalence group D of D^GThe accuracy is high, and before the equivalence relation judging method is utilized, special conditions that equivalence relation judgment is not established without inquiring SRS data are eliminated, SRS data inquiry is avoided, and the calculation efficiency is improved;

(4) step S5 of the present invention at X^GAfter non-empty set, traverse X^GMerging and D according to the equal-cost transfer assumption^GMaximum equivalence with intersectionGroup to maximum equivalence group D^GSo that and D^GSignals contained in the maximum equivalent group with intersection do not need to be subjected to equivalent judgment one by one, so that the number of relation judgment times in the equivalent group is reduced, the calculation amount is reduced, and the analysis efficiency is improved.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. An equivalent signal mining method for SRS combined adverse reaction signals is characterized by comprising the following steps:

s3, priority processing the signal with the smallest size of the drug combination set, i.e. order processing the signal

S8, order X^G＝X^G∪{D^GReverting to step S2;

s9, outputting the maximum equivalence group set X^GThe maximum equivalence group is a set of the maximum equivalence groups of the combined adverse reaction signals, each maximum equivalence group is a set of the combined adverse reaction signals with pairwise equivalence relation, and one signal, namely one drug combination, is selected from each maximum equivalence group, so that the maximum equivalence group can be represented to participate in the data analysis of the combined adverse reaction.

2. The SRS combination adverse drug reaction signal-oriented equivalent signal mining method according to claim 1, wherein the data is obtained from demographic information, medication information, adverse drug reaction information of a spontaneous report system SRS for adverse drug events.

3. The method for mining an equivalent signal for an adverse reaction signal to SRS combination according to claim 1, wherein the method for determining the drug combination for an adverse reaction signal comprises: if the frequency of the target adverse reaction signals AE reported by the patients taking the medicine combination D is greater than the support threshold; the length of the drug combination D is not more than the drug combination signal length threshold MAX _ D; d obtained from SRS data of PS patient set is used as lower bound Q (PS, D, AE) CI of signal intensity confidence interval of target adverse reaction signal AE^->A signal strength threshold δ; d is determined to be the drug combination of the targeted adverse reaction signal AE.

4. The method as claimed in claim 3, wherein Q (PS, D, AE) is a signal metric function of the non-equilibrium metric class, and the metric methods adopted include RR, PRR, RoR, BCPNN, MGPS.

5. The SRS combination adverse reaction signal-oriented equivalence signal mining method according to claim 1, wherein the preprocessing comprises de-duplicating data and normalizing drug names.

6. The method for mining an equivalent signal to an SRS combined adverse drug reaction signal according to claim 1, wherein the step S5 includes the steps of:

7. The method for mining an equivalent signal to an SRS combined adverse drug reaction signal according to claim 1, wherein the step S7 includes the steps of:

s7.1, selecting non-traversed member D of Z_i；

8. The SRS combination adverse reaction signal-oriented equivalent signal mining method according to claim 7, wherein in step S7.2, whether D is equivalent to D is judged_iThe method comprises the following steps:

s7.2.7, judging C (G), (D) -G (D)_i)，AE)<θ and C (G (D)_i)-G(D)，AE)<If θ is true, the routine proceeds to step S7.2.8; otherwise, go to step S7.3;

s7.2.8, D is equivalent to D_iLet D^G＝D^G∪{D_i}。

9. A server, comprising:

the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the equivalent signal mining method for SRS combination adverse reaction oriented signals as claimed in any one of claims 1 to 8.

10. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of equivalent signal mining for SRS combination adverse reaction signals according to any one of claims 1 to 8.