WellGreen-i Co., Ltd.

― WGI’s AI and Digital Transformation: Driving High-Quality Gene Expression Data for Rapid and Accurate Gene Identification ―

◆ In the following descriptions, the letter following each “Task” (e.g., Task A) corresponds to the codes (A–Q) listed under “Outsourcing task(s)” in the Contact Us form.

WGI’s proprietary technologies for high-precision and rapid gene discovery using gene expression data

WGI offers contract analysis services for gene identification and related purposes.
With extensive experience and a proven track record in gene discovery using gene expression data, we support researchers in achieving faster and more accurate results.
To accelerate gene discovery, WGI has developed proprietary AI-, DX-, and bioinformatics-based analysis platforms, combined with curated, high-quality big data resources.
These advanced analytical methods go beyond conventional techniques, enabling rapid and precise gene discovery while overcoming the limitations of traditional approaches.
Our system is capable of efficiently handling large-scale gene expression datasets.

Foundational Information for Transcriptome Analysis: Gene Expression Matrix and Gene Expression Profiles

◆Transcriptome Analysis and DEG Discovery from the Gene Expression Matrix

Transcriptome analysis is a key approach for identifying candidate genes associated with specific traits (including environmental responses and time-course experiments).
It enables comprehensive and high-precision analysis of genome-wide gene expression levels (i.e., transcriptome information).
By compiling expression levels obtained under multiple experimental conditions (including control experiments), a gene expression matrix is generated (see table below).

An Example of the Gene Expression Matrix (Unit: TPM)

◆Obtaining Gene Expression Profiles from the Gene Expression Matrix

Each row in the gene expression matrix (i.e., a row vector) represents the pattern of expression levels of a specific gene across all experimental conditions (runs) and can be visualized as a gene expression profile (see figure below).

An Example of the Gene Profile

By utilizing both the gene expression matrix and the gene expression profiles, various types of gene discovery become possible.
At WGI, we have established a cutting-edge proprietary transcriptome analysis platform to offer contract analysis services that rapidly and accurately derive biological insights from gene expression matrices.

High-Accuracy, High-Speed, and Cost-Effective Gene Discovery through Metadata Organization of RNA-Seq Big Data (Task D)

Inability to Fully Utilize RNA-seq Big Data Available Online

Extensive RNA-seq datasets are registered and publicly available in online databases, providing significant scientific value.
Leveraging these experimental datasets is expected to accelerate the identification of various genes.
However, because the description of RNA sampling conditions is provided by individual data submitters, the quality and level of detail in these descriptions vary widely across database records.

◆Generating High-Quality Big Data via Metadata Organization of Public RNA-seq Experimental Datasets

RNA-seq experimental conditions span a wide range of items such as material (strain/variety), growth stage, tissue type, time point, and cultivation environment.
It is time-consuming and labor-intensive to fully understand the experimental conditions of RNA-seq datasets obtained from databases.

At WGI, we investigate key experimental condition factors of RNA-seq experiments in model species registered in databases and organize RNA-seq experimental conditions into tabular format (metadata creation).
The metadata includes information such as experiment ID, material, treatment, growth stage, and sampling tissue (see the table below).
By utilizing metadata, it is possible to select a larger number of appropriate and relevant RNA-seq experimental datasets, thereby enabling big data analysis compared to standard database searches.
As a result, gene discovery becomes more accurate and efficient.
By combining gene expression matrices with metadata, it becomes easy to sort the X-axis of gene expression profiles by any experimental condition (e.g., sampling tissue, treatment, genetic background) using spreadsheet or statistical analysis software, thus accelerating gene discovery from expression profiles.
Metadata creation is performed by WGI’s experienced curators, and can be prepared for any species, including animals, plants, and microbes.

An Example of the RNA-Seq Metadata

◆Enhanced Accuracy and Speed of Gene Discovery through Ontology-Based Metadata

RNA-seq experimental conditions cover a wide range of factors, including material (strain/variety), growth stage, tissue, time point, and cultivation environment.
At WGI, for plant species, we utilize both Plant Ontology (PO, describing plant tissues) and Plant Experimental Conditions Ontology (PECO, describing treatments), enabling detailed classification and retrieval of RNA-seq datasets (see the table below).
For animals and microbes, we similarly create ontology-based metadata (standardized terminology) to enable comprehensive searches of RNA-seq datasets based on target experimental conditions.
By using large-scale RNA-seq metadata, it is possible to accurately and rapidly identify genes highly expressed under specific spatial, temporal, or environmental conditions.

An Example of the RNA-Seq Curation in a Plant Species

Generation of High-Quality Gene Expression Matrices (Task D, E)

At WGI, we use only high-quality sequencing reads filtered for DNA regions, ensuring that high-confidence data are applied in downstream analyses. Additionally, we handle only high-quality experimental datasets based on statistics such as mapping rates to reference genome sequences, enabling high-accuracy gene discovery.

◆Selection of High-Quality Experimental Data via Metadata, Generation of Gene Expression Matrices, and Various Analyses (Task D)

WGI’s contract services provide metadata and registered run count tables by experimental condition for the target species.
By selecting RNA-seq experimental data using this metadata, highly efficient and accurate analyses such as gene discovery are possible.
If you inform us in advance about tissues or treatments of interest for your target species, we can reduce the metadata volume, which is usually extensive, making the selection process easier.
WGI will generate and deliver gene expression matrices based on the finalized experimental datasets.
We also offer various analyses using gene expression matrices, including DEG selection, expression profile similarity/opposition analysis, functional annotation of candidate genes, and cross-species homolog (gene family) analysis.

◆Generation of Gene Expression Matrices from Proprietary RNA-Seq Data and Various Analyses (Task E)

We offer high-quality generation and provision of gene expression matrices from your proprietary RNA-seq data.
We can also integrate online RNA-seq data with your proprietary data to generate gene expression matrices (integration analysis with Task D).
Additionally, we offer various analyses using gene expression matrices, such as DEG selection, expression profile similarity/opposition analysis, functional annotation of candidate genes, and cross-species homolog (gene family) analysis.

High-Accuracy, Rapid, and Cost-Effective Gene Discovery from Gene Expression Matrices
(Flexible Analysis Options Combinable with Tasks D and E)

By leveraging gene expression matrices, it is possible to identify candidate target genes through various approaches.

◆Identification of Spatiotemporally Specific Expressed Genes and DEG Groups

Genes with spatiotemporally specific expression are potential candidates for master regulatory genes.

From gene expression matrices, it is possible to statistically and reliably identify genes with spatiotemporally specific expression, i.e., genes expressed predominantly under one or more RNA sampling conditions (e.g., tissue, growth environment).
For example, the figure below shows genes specifically highly expressed in flowers.
At WGI, by creating and utilizing RNA-seq experimental metadata, it is possible to collect the maximum number of experimental datasets under the same conditions and treat them as biological replicates.
Therefore, compared to exploring target genes from a single control experimental dataset, it is possible to extract target gene groups with higher reproducibility and accuracy.
Furthermore, conventional DEG screening methods face limitations in sufficiently narrowing down candidate gene groups.
Therefore, at WGI, we utilize our proprietary AI analysis platform for high-accuracy DEG extraction to identify genes exhibiting specific expression profiles.

At WGI, to accurately identify DEG groups and genes with spatiotemporally specific expression, we utilize not only metadata but also our proprietary statistical analysis platform.
As a result, we can obtain high-accuracy candidate gene information that is not achievable through conventional approaches relying on standard statistical methods and database-sourced experimental datasets.

An Example of a Specifically Expressed Genes

◆High-Efficiency Selection of Candidate Genes Sharing Identical or Similar Biological Processes and Expression Regulatory Mechanisms (Task F)

Gene groups with similar expression profiles (figure below) tend to be up- or down-regulated under the same spatiotemporal and environmental conditions, indicating biological relevance.
For example, such similarity suggests involvement in a common biological process (e.g., gametogenesis, a specific metabolic pathway) or regulation by identical or related transcription factors and cis-elements.
In cases where one aims to discover genes co-acting with a known gene, genes with expression profiles similar to the known gene serve as candidate gene groups.
From experimental conditions with high expression levels, one can infer the spatiotemporal and treatment conditions in which the genes are expressed and functional.

At WGI, we have established and utilize a proprietary AI/DX analysis platform to identify genes with similar expression profiles in a short time, with high precision and low cost.
Compared to conventional correlation-based co-expression methods between two genes, our approach avoids false positives, reduces computation time and memory usage, and addresses the limitations in accuracy and efficiency.

An Example of Similar Gene Expression Profiling

◆High-Efficiency Discovery of Gene Groups Involved in Negative Feedback Mechanisms or Enzyme Genes Acting on the Same Substrate (Task F)

Gene groups with reciprocal (opposite) expression profiles (figure below) are presumed to be related via negative feedback mechanisms.
For instance, if genes encode enzymes acting on the same metabolic precursor, higher activity (higher expression) of one enzyme may preferentially consume the precursor, indicating a trade-off relationship in precursor utilization.
Such gene groups may show reciprocal expression profiles.
Information on reciprocal expression profiles is useful for elucidating negative feedback mechanisms and identifying enzyme gene groups acting on the same precursor.
By examining experimental conditions associated with high and low expression, it is possible to infer control conditions governing negative feedback or trade-offs.

At WGI, we have established and utilize a proprietary AI/DX analysis platform to identify genes with reciprocal expression profiles in a short time, with high precision and low cost.
Compared to the correlation-based methods (co-expression analysis) between two genes used in conventional analysis, our approach avoids false positives and reduces computation time and memory usage, overcoming issues of accuracy and efficiency.

An Example of Reciprocal Gene Expression Profiling

◆Identification of Gene Groups with Similar and Reciprocal Expression Profiles by Conventional Methods

Correlation analysis (correlation coefficients) has conventionally been used to identify gene groups with similar or reciprocal expression profiles.
While correlation analysis is not recommended as the primary method due to the risk of false positives, it can be used as reference information.
Besides correlation coefficients, other metrics such as Euclidean distance can be used, but they require high computational resources (time and memory), especially with large datasets.

◆Annotation with Biological Function Knowledge of Genes (Task B, C)

Function prediction based on DNA/amino acid sequence similarity, which is widely used, lacks biological validation and is therefore not suitable as a primary approach in gene discovery.
Enrichment analysis based on functional annotation of candidate gene groups also has statistical limitations.
While sequence similarity analysis and enrichment analysis can be used as reference information, their lack of biological basis often reduces the efficiency of gene discovery.
In addition to such reference information, WGI can provide high-confidence functional knowledge obtained through our proprietary AI text-mining technologies (A2K/LA2K).
Through A2K/LA2K analysis of peer-reviewed literature abstracts and other texts, knowledge regarding gene function, metabolism, expression regulation, and environmental responses can be extracted.

◆Annotation with Protein Functional Domain Information and Intra-/Inter-Species Homolog (Gene Family) Information (Task G, H)

Because sequence similarity-based function prediction contains many false positives, WGI additionally provides more reliable domain-level functional annotations (Task G).
WGI performs gene family classification based on domain and sequence pattern similarities, which provides insights into shared biological functions and regulatory mechanisms (Task H).

◆Annotation with Gene Expression Regulators: Transcription Factors and Cis-Elements (Task B, C, I)

WGI collects gene regulatory information using proprietary AI text-mining methods (Task B, C) and AI-driven sequence analysis techniques (Task I).

◆Gene–Compound and Other Network Construction

Relationships between genes and compounds (such as metabolites) are visualized through network graphs.
By using free network visualization and editing tools, users can rapidly and easily interpret gene–compound relationships, facilitating functional inference for uncharacterized genes.
Relationship types include DNA/amino acid sequence similarity, expression profile similarity/reciprocity, spatiotemporal high expression patterns, associated traits/metabolic pathways, transcription factors, and intra-/inter-species homologs.
For intra-/inter-species homologs, WGI can annotate corresponding information, enabling cross-species utilization of gene-related knowledge and maximizing the information available for gene discovery.