AI創薬支援

AI Drug Discovery
Support Platform

Drug Discovery AI Factory

Hypothesis generation for drug discovery

AI “KIBIT”, 
Natural Language
and Biologists
generate Hypothesis
as a starting point for drug discovery

AI KIBIT AI KIBIT

Features of Drug Discovery AI Factory

Novelty

Discovery of unreported target molecules with high disease relevance utilizing self-developed AI “KIBIT”

High accuracy

Unique analysis methods extract target molecules with high probability of success

Hypothesis generation

Create gene network and generate hypothesis for unreported target molecules

Extraction of many unreported target molecules by utilizing KIBITNarrowing down the target molecules to even more precise targets utilizing original analysis methods

KIBIT is our proprietary natural language processing AI engine, designed to discover unreported relationships — associations not documented in literature — through a novel approach called non-continuous discovery. At the Drug Discovery AI Factory, we utilize KIBIT to generate networks that analyze the relationships among molecules associated with a disease. In this process, KIBIT also predicts target molecules even if their relationships to the disease are not reported.
Based on the network, we utilize our own innovative analysis methods, the Drug Discovery Best Known Method (DD-BKM), to narrow down the targets with high potential for new drugs.

Example: Gene network for disease A

ddaif_img3: Target molecules that are highly relevant to disease A and whose relevance is not described in a paper

map_ex1

 

Disiease A Gene B
icon_search
icon_down_cyan
No hit
Disiease A Gene C
icon_search
icon_down_cyan
No hit

No linkage to disease A
has been reported for both gene B and gene C

Extraction of genes with high probability of drug discovery success

ddaif_img3: Target molecules with a high probability of successful drug discovery are selected from among target molecules that are highly relevant to disease A and whose relevance has not been described in papers.

map_ex2

 

Analysis example

Virtual Experiments with KIBIT
（Virtual Experiments）

We simulate how pathways are changed by hypothetically knocking out (KO) the target gene. For example, in the figure below, a hypothetical KO of gene B has almost no effect on the network as it is only replaced by gene D. On the other hand, when gene C is knocked out, the network changes significantly, suggesting that gene C is the gene affecting the disease.

ddaif_img2

 

Unique technology to discover unknowns from known informationPredicts even molecules that have no association with disease in literatures

The main source of article data for analysis is PubMed, and Springer Nature article data can be added as an option. FRONTEO's drug discovery support platform is characterized by its ability to extract novel target molecules from previously reported information.

KIBIT can read known article information and derive unreported relationships.

ddaif_img4

 

Springer Nature

Extended function
Analyzes full-text data of about 600 journals published by Springer Nature
Promoting innovative discoveries based on a variety of cross-disciplinary findings by analyzing data from biomedical, biotechnology, and other fields

Comparison of full text analysis and abstract analysis

ddaif_img7

Increasing the number of genes analyzed by KIBIT allows for the
discovery of more novel target molecules.

ddaif_img6

Bridges the time lag between being reported in abstract and discovering
target molecules on average 5 years earlier

Click here for more information

Experts with years of experience in drug discovery research make proposals that meet customer needs

To analyze highly specialized information and derive innovative outputs, drug discovery researchers with expertise in both drug discovery and AI is essential. FRONTEO brings together a team of researchers who have long been engaged in drug discovery at pharmaceutical companies and research institutions to realize useful proposals, including appropriate target molecules and indications and their hypotheses, in line with the needs of our clients.

staff1

Hiroyoshi Toyoshiba

Director/CTO
Doctor of Science

Graduated from Waseda University Graduate School of Science and Engineering Department of Mathematics. Engaged in research using statistical analysis of various data at the US National Institute of Environmental Health Sciences, and others. Developed AI algorithms for life sciences.
staff2

Makoto Miyamoto

Research Team Director
Doctor of Agriculture

Graduated from Kyoto University graduate school. At Takeda Pharmaceutical Company Limited, he was engaged extensively in preclinical safety evaluation, special toxicity (phototoxicity) evaluation, elucidation of toxicity mechanisms, and safety biomarker searches from the early to late stages of new drug development.
staff3

Kazumi Hayashi

Research Team
Doctor of Pharmacy

After receiving his doctorate in pharmacy, he joined Yamanouchi Pharmaceutical (currently Astellas Pharma). As a research manager, he is in charge of inflammation, cardiovascular, and renal areas. He set up a drug repurposing room and engaged in DR research. He is in charge of devising analysis methods using KIBIT, planning analysis policies, analysis, and generating hypotheses.

FRONTEOの考える仮説

ddaif_img8

The hypothesis to initiate the development of a new drug includes

  1. Target molecule: A gene or molecule etc., that is predicted to be associated with a disease.
  2. Disease mechanism: How the target molecule is related to the disease, etc.
  3. Patient information: Genome information, diseases, symptoms, etc., of the target patient group.
  4. Safety information: Toxicity of the drug candidate, etc.
  5. Feasibility: Proposed experimental model, etc.

FRONTEO believes that all of these factors need to be considered.

  • HOME      >      
  • AI Drug Discovery Support Platform