LEXAS - Lifescience experiment search and suggestion -

Overview

LEXAS is a text-mining-based web application that helps life science researchers to efficiently design biological experiments.
With LEXAS, you can quickly find biological experiments and receive suggestions for the next experiments from machine learning models.

Search

The LEXAS "search" interface allows you to quickly find descriptions about biological experiments in the literature.
In LEXAS, the pairs of a gene name and an experiment name mentioned in the same sentence are filtered by a fine-tuned bio-BERT model and extracted as descriptions of biological experiments. The sentences were retrieved from the result section of all PMC (PubMed Central) articles (last update:2021/9/5).
The list of gene names used for this retrieval consists of 106,956 terms provided by the HUGO Gene Nomenclature Committee (HGNC).
The list of terms of experiment methods used for this retrieval is available on this page.

How to use

1. Enter a gene name and select an experimental method

2. The matching experiments are shown

How to read the result table

Confidence
The confidence score given by the machine learning model, which indicates the likelihood that the extraction is correct.
Year
The year of publication.
PMCID
The PMCID of the article.
Gene
The notation of the gene name in the sentence.
Experiment
The notation of the experiment in the sentence.
Sentence
The sentence describing the experiment. The genes and methods are marked with blue and yellow lines, respectively.
License
License of the article that describes the diplayed sentence.

Suggestion

The LEXAS "suggestion" interface shows a ranked list of genes that are predicted to be plausible targets for the next biological experiments by a machine learning model.
LEXAS also suggests possible experiment methods to analyze the function of the suggested gene.

How to use

1. Enter the gene name that you have analyzed experimentally and select an information source

2. The genes to be analyzed in the next experiment are shown.

How to read the result table

Rank
The rank of the suggested genes.
Symbol
Gene symbol
Score
The confidence score given by the machine learning model, which indicates the likelihood that the suggestion is correct.
Features and their importance
(XXXX, 0.78) means that the feature XXXX is shared by the query and output genes, and its importance (SHAP value) is 0.78.
Mentioned in the same paper
Whether the query gene and the output gene are mentioned in the same published article.
Experiment
Possible experimental methods to analyze the suggested genes.

List of information sources

In LEXAS, we refer to an information source annotated manually by trained researchers or medical doctors as a knowledgebase, and an objective information source such as expression levels as a database.

Information Source	Information type	Description	Term in the result table
Gene Ontology	Knowledgebase	Biological process, molecular function, and cellular component	GO: XXXXX
Mouse genome informatics	Database	Phenotypes of knockout mice	MP: XXXXX
Human phenotype ontology	Database	Phenotypic abnormalities in human	HP: XXXXX
Online Mendelian Inheritance in Man	Knowledgebase	Genetic diseases	OMIM: XXXXX
Orphanet	Knowledgebase	Genetic diseases	ORPHA: XXXXX
Human protein atlas (1)	Database	Subcellular localization	(A name of cellular component)
Human protein atlas (2)	Database	Expression levels among human tissues	(A name of tissue)-(H/L)-(Level)*
BioGRID	Database	Protein-protein interaction	BG: XXXXX
DepMap	Database	Cancer cell growth under CRISPR/Cas9 mediated suppression of genes	DepMap
ENCODE	Database	Transcription factor	TF: XXXXX
Word2Vec	Knowledgebase	The similarity of the usage of gene terms in the MEDLINE abstract	Word2Vec

* (example) Testis-H5: Expression level of a gene in testis is much higher than those in other tissues.
* (example) Testis-H1: Expression level of a gene in testis is a little higher than those in other tissues.

Contributors

Graduate School of Pharmaceutical Sciences, The University of Tokyo

Kei K Ito
Daiju Kitagawa

Graduate School of Information Science and Technology, The University of Tokyo

Yoshimasa Tsuruoka

Contact/Feedback

lexas.f.utokyo [at] gmail.com

Version

Version	Date	Content
1.06	current: Nobember 11, 2024	Data updated
1.05	February 8, 2024	Add license column
1.04	June 24, 2023	Data updated; free text search bocomes available
1.03	January 20, 2023	Set up a feedback page
1.02	October 18, 2022	Data updated
1.01	December 3, 2021	Examples of usage are posted on the website.
1.0	October 25, 2021	20120 genes in 1 organism (human)

Acknowledgement

LEXAS relies on many information sources:

PubMed central HGNC GO MGI HPO OMIM Orphanet HPA BioGRID DepMap ENCODE STRING

We gratefully acknowledge these resources.

Software

LEXAS is built using Python version 3.6.9

Citation

Predicting potential target genes in molecular biology experiments using machine learning and multifaceted data sources
Kei K Ito, Yoshimasa Tsuruoka, Daiju Kitagawa
iScience Volume 27, Issue 3, 15 March 2024, 109309
doi:https://doi.org/10.1016/j.isci.2024.109309

LEXAS: a web application for life science experiment search and suggestion
Kei K Ito, Yoshimasa Tsuruoka, Daiju Kitagawa
bioRxiv 2021.12.05.471323;
doi: https://doi.org/10.1101/2021.12.05.471323

Licensing

Search

All data and download files available through LEXAS "Search" are licensed under the terms specified in the "License" column.

When using the LEXAS "Search" data, please indicate proper credit.

Suggestion

All data and download files in LEXAS "Suggestion" are licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

When using the LEXAS "Suggestion" data, please indicate proper credit and inform the users of any modifications that you added to the data.

Source code

The source code may be used for non-commercial purposes. If you want to use the source code for a commercial purpose, please contact the following email addresses.
Miho Sakao : sakao [at_mark] todaitlo.jp

LEXAS - Lifescience experiment search and suggestion -

Overview

Search

How to use

How to read the result table

Confidence

Year

PMCID

Gene

Experiment

Sentence

License

Suggestion

How to use

How to read the result table

Rank

Symbol

Score

Features and their importance

Mentioned in the same paper

Experiment

List of information sources

Contributors

Graduate School of Pharmaceutical Sciences, The University of Tokyo

Graduate School of Information Science and Technology, The University of Tokyo

Contact/Feedback

Version

Acknowledgement

Software

Citation

Licensing

Search

Suggestion

Source code