Back

Natural language processing of gene descriptions for overrepresentation analysis with GeneTEA

Boyle, I. A.; Aquib, N. A.; Kocak, M.; Creasi, R.; Montgomery, P. G.; Campbell, C. D.; Dempster, J. M.

2025-04-03 bioinformatics
10.1101/2025.03.28.646026 bioRxiv
Show abstract

Overrepresentation analysis (ORA) is used to identify the biological relationships in a list of genes by testing gene sets for enrichment in the query. However, the inconsistent definition and highly overlapping nature of gene set databases can make interpreting ORA results difficult. Here, we introduce GeneTEA, a model that takes in free-text gene descriptions and incorporates several natural language processing methods to learn a sparse gene-by-term embedding, which can be treated as a de novo gene set database. We benchmark performance against other popular ORA tools and find that only GeneTEA properly controls false discovery while consistently surfacing relevant biology. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=75 SRC="FIGDIR/small/646026v1_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@145a76aorg.highwire.dtl.DTLVardef@1f221d3org.highwire.dtl.DTLVardef@18acdacorg.highwire.dtl.DTLVardef@1c5106a_HPS_FORMAT_FIGEXP M_FIG C_FIG

Published in Genome Biology (predicted rank #3) · training set

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

1
Nucleic Acids Research
◐ 1281 papers in training set
Top 0.9%
17.8%
Similar papers in this journal
2
NAR Genomics and Bioinformatics
◐ 242 papers in training set
Top 0.1%
14.5%
Similar papers in this journal
Genome Biology · published here
○ 637 papers in training set
Top 1%
9.4%
Similar papers in this journal
4
Bioinformatics
◐ 1204 papers in training set
Top 3%
9.4%
Similar papers in this journal
50% of probability mass above
5
Nature Methods
○ 385 papers in training set
Top 2%
5.3%
Similar papers in this journal
6
Bioinformatics Advances
◐ 203 papers in training set
Top 1%
5.0%
7
PLOS Computational Biology
● 1863 papers in training set
Top 8%
4.2%
8
Cell Genomics
○ 172 papers in training set
Top 1%
4.2%
9
Briefings in Bioinformatics
◐ 354 papers in training set
Top 2%
4.2%
10
Proceedings of the National Academy of Sciences
● 2444 papers in training set
Top 18%
3.1%
11
Cell Systems
○ 201 papers in training set
Top 2%
2.3%
12
Nature Communications
○ 5641 papers in training set
Top 42%
2.0%
13
Science Advances
● 1243 papers in training set
Top 21%
1.7%
14
Patterns
○ 78 papers in training set
Top 1%
1.6%
15
Science
● 477 papers in training set
Top 7%
1.3%
16
Computational and Structural Biotechnology Journal
● 242 papers in training set
Top 4%
1.3%
17
eLife
● 5828 papers in training set
Top 61%
1.1%
18
BMC Bioinformatics
○ 457 papers in training set
Top 5%
1.0%
19
Nature Computational Science
○ 55 papers in training set
Top 2%
0.8%
20
Molecular Systems Biology
○ 162 papers in training set
Top 3%
0.8%
21
Cell Reports Methods
○ 165 papers in training set
Top 5%
0.6%
22
Genome Research
● 468 papers in training set
Top 7%
0.6%

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.