D2H2


Chatbot	Single Gene Queries	Gene Set Queries	Bulk Studies	Single Cell Studies	Hypotheses	Resources	Contribute	Downloads	About	Help

D2H2 Automated Hypothesis Each and Every Day (AHEAED)

To form novel hypotheses, each day we query the D2H2 gene sets created from the collection of bulk RNA-seq studies against >600,000 gene sets extracted from the supporting materials of research publications listed in PubMed Central (see rummagene.com). We then identify studies that have high overlap with the D2H2 gene set and low abstract similarity. We then pass the two dissimilar abstracts to GPT4 for hypothesis generation. The LLM model attempts to reason about possible explanation of why the studies with such dissimilar abstracts have such similar gene sets.

Search Previous Hypotheses:

Try an example: pancreas, aging, adipose

Today's Hypothesis

**These hypotheses are generated using GPT4 and abstracts of the corresponding studies. Please use caution when interpreting the results.**

Find studies with similar gene sets to your input gene set but with dissimilar abstracts

0 gene(s) entered

Try an example gene set

File formats accepted: csv, tsv, txt file with Entrez gene symbols on each line

Enter gene set description (optional):

Try an example abstract

Perform enrichment analysis on hundreds of thousands gene sets extracted from supporting tables of over one hundred thousand articles to find the most similar gene sets that match your query. The top 100 are returned and are ranked by cosine similarity based upon your entered abstract and those of the enriched gene sets.