Back

GenePT Revisited: Do Better Text Embeddings Make Better Gene Embeddings?

Hedley, J. G.; Torr, P. H. S.; Märtens, K.

2026-04-20 genomics
10.64898/2026.04.16.718976 bioRxiv
Show abstract

AO_SCPLOWBSTRACTC_SCPLOWGenePT introduced a simple recipe for gene representations: embed each genes natural-language description with a general-purpose text embedding model and reuse the resulting vectors across downstream tasks. Since GenePTs release, embedding models have improved rapidly, with many strong open and commercial encoders benchmarked on suites such as the Massive Text Embedding Benchmark (MTEB). We present a controlled "leaderboard" study that keeps the GenePT pipeline fixed and varies only the embedding backbone. We benchmark contemporary encoders on four diverse gene embedding tasks: gene-gene interaction prediction, gene property classification, cell type classification, and prediction of transcriptomic responses to unseen genetic perturbations. Across these settings, newer backbones consistently outperform the original GenePT backbone (text-embedding-ada-002), achieving improvements of 1-17%, while enabling fully reproducible research by avoiding API dependencies.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

1
Genome Research
409 papers in training set
Top 0.1%
12.6%
2
Genome Biology
555 papers in training set
Top 0.4%
10.0%
3
Nature Biotechnology
147 papers in training set
Top 0.7%
10.0%
4
Nature Methods
336 papers in training set
Top 2%
6.3%
5
Nature
575 papers in training set
Top 4%
6.3%
6
Nature Machine Intelligence
61 papers in training set
Top 0.6%
4.8%
7
Nature Communications
4913 papers in training set
Top 33%
4.8%
50% of probability mass above
8
Bioinformatics
1061 papers in training set
Top 4%
4.8%
9
Nucleic Acids Research
1128 papers in training set
Top 5%
4.1%
10
Nature Genetics
240 papers in training set
Top 2%
3.6%
11
Bioinformatics Advances
184 papers in training set
Top 2%
2.8%
12
Cell Genomics
162 papers in training set
Top 2%
2.4%
13
Cell Systems
167 papers in training set
Top 6%
2.1%
14
Nature Computational Science
50 papers in training set
Top 0.4%
2.1%
15
Proceedings of the National Academy of Sciences
2130 papers in training set
Top 33%
1.7%
16
Science
429 papers in training set
Top 14%
1.7%
17
Genome Medicine
154 papers in training set
Top 6%
1.3%
18
Frontiers in Genetics
197 papers in training set
Top 6%
1.3%
19
Nature Medicine
117 papers in training set
Top 3%
1.2%
20
Briefings in Bioinformatics
326 papers in training set
Top 5%
0.9%
21
Scientific Reports
3102 papers in training set
Top 71%
0.9%
22
Nature Neuroscience
216 papers in training set
Top 5%
0.9%
23
IEEE Transactions on Computational Biology and Bioinformatics
17 papers in training set
Top 0.7%
0.7%
24
BMC Bioinformatics
383 papers in training set
Top 7%
0.7%
25
PLOS ONE
4510 papers in training set
Top 68%
0.7%
26
The American Journal of Human Genetics
206 papers in training set
Top 4%
0.7%