Welcome to the SKiM Suite

A two-stage tool for biomedical literature research: find compelling term co-occurrence results in 40M+ PubMed abstracts, then evaluate or compare the hypotheses they suggest with a large language model. Free to use — just an email address to register.

Stage 1

Co-occurrence search

Rank pairs (or chains) of biomedical terms by how strongly they co-occur across PubMed abstracts, using Fisher’s Exact Test or Chi-square. A strong signal is a starting point worth investigating — not a causal claim.

Stage 2

Hypothesis testing

Slot co-occurrence results into a hypothesis template (e.g. “{A} informs {B}”) and put them to the test with an LLM: evaluate how well the literature supports a single claim, or compare two competing claims head-to-head.

Start a co-occurrence search

Choose how many terms your search spans.

2-term A → B 3-term A → B → C

Rank B-terms by their strength of direct co-occurrence with an A-term in PubMed abstracts. Best for prioritizing research directions when you want to know which of many candidates is most connected to a concept you care about.

Example hypotheses this might surface:

metformin is associated with longevity
Alzheimer’s disease is linked to tau protein

Start a 2-term search

If you use the SKiM Suite in your research, please cite:

2-term co-occurrence search (KinderMiner): https://pmc.ncbi.nlm.nih.gov/articles/PMC5543342/
3-term co-occurrence search (SKiM): https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-023-05539-y
Hypothesis evaluation: https://link.springer.com/article/10.1186/s12859-025-06350-7
Direct comparison of hypotheses: {pre print here}