How much data is stored in the compound database?
As of February 2021, the number of existing compounds is approximately 107,000,000; the number of new chemical compounds is approximately 11,000,000; The number of patents is 25,000,000; the number of papers is 32,000,000; the number of characteristic data is 28,200,000.
We are planning to update the compound database once every six months.
What patents are supported?
Patents are registered in 105 countries and regions including US, EP, CN, JP, KR, TW and WO, etc.
Is it possible to import our company's data into Chemicals Informatics?
At the moment, the NLP API is not open to the public. Basically, we only handle publicly available data.
How many different properties are there?
As of February 2021, the database includes information on 36 types of material properties and biochemical properties. It is also possible to output the predicted value concerning each property value.
Will the number of properties increase?
We are adding them as they develop. In addition, we may be able to add the requested properties by updating the product. Please consult with us for further information.
What is the principle of "cross-search" used by search AI?
We adopt a computational method that emphasizes explanatory power, in which features of organic structures and inorganic elements of about 100 dimensions are converted into vectors, and the Euclidean distances between feature vectors are calculated. It is also possible to add weights to each feature.
Does cross-search mean to search for a combined or mixed material that is a combination of A and B? Or does it mean to search for a structure that is close to A and B?
Both patterns can be explored simultaneously. Not only two systems, A and B, but also up to 4 systems of crossings for mixtures, and up to 64 systems of crossings for structures can be searched. The compound shows patents that commonly include compounds of all specified systems. (As of February 2021)
Are both organic and inorganic compounds supported?
Yes, both are supported.
Does the system support polymers as well as monomers?
Yes, it is supported. It is also possible to specify the degree of polymerization in the search conditions.
Is it possible to search by targeted properties?
Yes, it is possible. If you specify only a small number of properties as search conditions, the number of compounds displayed as search results is likely to be huge; we recommend that you specify multiple properties or specify compounds with properties close to the target properties at the same time.
Why is it possible to predict properties using the open literature? How accurate is it?
By predicting properties based on the number of literature distributions that consider the applications of similar compound groups, 15 properties can be predicted with correlation coefficients of 0.71 to 0.95 as of February 2021. Compounds and literature used in the prediction can be confirmed in full English translation from the literature links categorized by property value. Detailed preconditions and production methods can be confirmed for each property value.
Has Hitachi High-Tech Solutions obtained patents for the compounds generated by the new compound generation AI?
No, we do not file patent applications. If you confirm the efficacy of a compound generated by our new compound generation AI, you may file a patent application.