JPH0646180B2 - Method for estimating partial structure of compound - Google Patents

Description

TECHNICAL FIELD The present invention relates to a method for estimating a partial structure of an unknown compound by applying an artificial intelligence method to infrared spectrum data of the unknown compound.

[Conventional technology]

There are two methods for determining the structure of an unknown compound from spectral information such as infrared spectrum, an inductive method and a deductive method. The inductive method is to create a database of spectra for many known compounds,
The search is to find a spectrum similar to the spectrum of an unknown compound. This method is called so-called file search or library search.
Tanabe, Special Issue on Chemistry, 140, p.107 (1983), JP-A-63
-111429, etc. On the other hand, the deductive method is to analyze the spectrum to obtain information on the partial structure of the unknown compound, and combine them to estimate the overall structure of the unknown compound. The present inventors have analyzed the chemical for those using the technique of pattern recognition of this method, 35, 518 (1986), it has been proposed in the 36, 339 (1987).

[Problems to be Solved by the Invention]

However, the above conventional method has the following problems.

The inductive method has a problem that the search data cannot be obtained unless the spectral data of the compound to be searched is stored in the database file in advance. Therefore, the application range is limited.

On the other hand, in the case of the deductive method, there is no limitation on the applicable range as described above, but in the method using the pattern recognition method, since the spectral data is processed as a pattern figure, many of the methods used for partial structure estimation are used. There is a problem that it is difficult to manage rules.

In view of the above points, the present invention, by adopting an artificial intelligence method in the deductive method, there is no limit to the scope of application, it is very easy to manage a large number of rules, a partial structure estimation method of the compound, The purpose is to provide.

[Means for Solving the Problems]

In order to achieve the above object, the method for estimating the partial structure of a compound of the present invention is to extract infrared spectrum data of a compound having the same partial structure from a large number of known compounds and average them with an absorbance scale. The relationship between the characteristic wavenumber and intensity data of the spectrum and the partial structure is extracted as a rule, and when the rule is extracted, the standard deviation of the wavenumber and intensity data of the peak is set as an allowable range. Means for storing the extracted and extracted rules as a knowledge base (FIG. 1), means b for inputting the wave number and intensity data of the characteristic peak of the infrared spectrum of an unknown compound, and the input data For the above rules,
Means c for determining the possibility of the existence of each partial structure, and means d for calculating the certainty factor of the existence of the partial structure whose existence is suggested using the input data and the rule description data.
It is characterized in that it is provided with a partial structure whose existence is suggested and a means e for outputting a certainty factor of the existence.

[Work]

In the present invention, based on the infrared spectrum data of a large number of known compounds, the relationship between the wave number and intensity of the peak of the infrared spectrum and the partial structure is extracted as a rule, and these rules are stored as a knowledge base. Every time, since the artificial intelligence method of estimating the partial structure by applying the above rule to the data of the wave number and intensity of the characteristic peak of the infrared spectrum of the unknown compound is used, the correct answer is the same as the inductive method. Unless there is no answer without it, the degree of certainty indicates the possibility of the existence of each substructure, which makes it possible to limit the scope of application. Moreover, the artificial intelligence approach makes it very easy to manage the large number of rules used for estimation. Therefore, the above problem is solved.

〔Example〕

 Hereinafter, the present invention will be described in detail with reference to examples.

FIG. 1 is a flow chart showing the procedure of the method of the present embodiment,
FIG. 2 is a block diagram showing a configuration of an expert system used for implementing the present embodiment. The expert system uses a so-called production room type knowledge representation, and basically has a user interface 1 and a knowledge base management. A system 2, a knowledge acquisition mechanism 3, an inference / certainty factor calculation unit 4, and a knowledge base 5 are provided. A user interface 1 is an interface between a user and an expert system, which inputs infrared spectrum data of an unknown compound, various commands, and outputs inference results. The knowledge base management system 2 manages the inference / certainty factor calculation unit 4 and the knowledge base 5 which are the core of this expert system. Knowledge acquisition mechanism 3 is a knowledge possessed by an expert, that is, knowledge is obtained by extracting a rule indicating a relationship between an infrared spectrum and a partial structure from infrared spectrum data of many known compounds accumulated in SDBS (spectral database system). Store in base 5. The inference / confidence calculation unit 4 sequentially applies the rules stored in the knowledge base 5 to the input data from the user interface to estimate the possibility of each substructure,
The certainty factor of existence is calculated for the partial structure in which the possibility of existence is suggested. Knowledge base 5 stores the infrared spectrum as a rule in relation to the substructure, and uses it for substructure estimation.

Next, the specific procedure of the method of this embodiment will be sequentially described with reference to the flowchart of FIG.

First, rules are extracted through the knowledge acquisition mechanism 3 based on the infrared spectrum data of known compounds and stored in the knowledge base 5 (step). In this example, rules were extracted based on infrared spectrum data of about 13,000 kinds of organic compounds accumulated in a spectrum database system (SDBS). That is, the infrared spectrum of a compound having the same partial structure (for example, saturated aldehyde) was extracted, averaged on the absorbance scale, and the wave number and intensity (relative absorbance) of the characteristic peak were taken as a rule. At that time, the standard deviation of the wave number and intensity of the peak was obtained by statistical processing, and they were set as the allowable range and included in the rule. It should be noted that data such as peak shape and width can also be effective rules when estimating the partial structure. However, it is not easy to express such data as a rule, and it is considered that it is not so easy for a novice user to input such data.
Here, I decided to use only the wave number and intensity of the peak as a rule. As the expert system construction tool, E-shell of Fujitsu Limited was used, and as the hardware, FACOM-M380 was used. The rules extracted as described above for various substructures are stored in the E shell as so-called production rules in the form of IF ... THEN ....

Next, the wave number and intensity of the main peak in the infrared spectrum data of the unknown compound to be estimated are input from the user interface 1 as input data (step). The peak intensity data may be transmittance or absorbance, but in the case of transmittance, it is converted to absorbance after input and converted into a numerical value of relative intensity.

When the input of data is completed, the inference / certainty factor calculation unit 4 applies the rules stored in the knowledge base 5 to determine the possibility of existence of each substructure (step). At this time, the data of the peak wave number and the allowable range of the intensity described in the rule are utilized. Next, the inference / certainty factor calculation unit 4 calculates the certainty factor of the existence of the partial structure using the input data of the wave number and intensity of each peak and the rule description data for the partial structure whose existence is suggested (step ), And memorize the result for one day (step). The above inference and confidence calculation are performed for all structures (step → step →
Step → step → step…).

When the inference and the confidence factor calculation are completed for all the substructures, the substructures whose existence is suggested and the certainty factor of the existence are output by the user interface 1 and displayed to the user.

Next, a description will be given of a test performed by using the measured spectrum for several kinds of partial structures using the above method.

For partial structures such as saturated aldehydes, saturated esters, saturated carboxylic acids, saturated ethers, nitro groups, cyano groups, etc., the rules were obtained from SDBS, the execution test using the measured spectrum was performed, and the confidence level output was It was verified.

As a result, it was found that the substructures used in the tests could all exist with high confidence.

〔The invention's effect〕

As described above in detail, according to the present invention, the relation between the data regarding the characteristic peak of the infrared spectrum of the known compound and the partial structure is extracted as a rule and is described as a knowledge base. Since the rules are applied to determine the possibility of the existence of each substructure and the degree of certainty of the existence of the substructure in which the existence is suggested is obtained, the substructure is estimated. It has the advantage that there are no restrictions on the scope of application and that a large number of rules can be managed very easily.

[Brief description of drawings]

FIG. 1 is a flow chart showing a specific procedure of a method of an embodiment of the present invention, and FIG. 2 is a block diagram showing a configuration of an expert system used for implementing the above embodiment. 1 …… User interface 2 …… Knowledge base management system 3 …… Knowledge acquisition mechanism 4 …… Inference / certainty factor calculation unit 5 …… Knowledge base

Claims (1)

[Claims] 1. Infrared spectrum data of a compound having the same partial structure is extracted from a large number of known compounds, and they are averaged on an absorbance scale to obtain data of wave numbers and intensities of characteristic peaks of the spectrum. The relationship with the substructure is extracted as a rule, and when the rule is extracted, the standard deviation of the wave number and intensity data of the peak is included in the rule as an allowable range, and the extracted rule is stored as a knowledge base. Means, means for inputting wave number and intensity data of a characteristic peak of the infrared spectrum of an unknown compound, and means for applying the above rule to the input data and determining the possibility of existence of each partial structure. , Means for calculating the certainty factor of existence of the partial structure of which existence is suggested using the input data and rule description data, and the partial structure of which existence is suggested and its A method for estimating a partial structure of a compound, comprising: a means for outputting a certainty factor of existence.