About Hot/Cold Diagnosis
A disease guessing game for medical students, built on a published network of 4,219 diseases and 322 symptoms.
The data
Zhou X, Menche J, Barabási A-L, Sharma A. Human symptoms–disease network. Nature Communications 5, 4212 (2014). doi.org/10.1038/ncomms5212
This game would not exist without that paper. The authors mined PubMed records and MeSH terms to link symptoms with diseases, and shared the results as supplementary data. Thank you to them.
How the game uses the data
Warmth comes from how diseases appear together with symptoms in the literature up to 2014. It is a study aid, not a clinical judgement, so some readings will surprise you.
Compares symptom profiles
Each disease has a weighted symptom profile. Similarity is the cosine between two profiles, the method the paper uses, with PubMed counts damped so one frequent term can't dominate.
Reduces generic symptoms
Fatigue, headache, nausea, fever and similar symptoms count for a quarter of their usual weight. A few terms that aren't clues, such as body weight, are left out.
Turns rank into warmth
Top scores vary between diseases, so warmth uses rank. The closest 1% of diseases are hot and the next 9% are warm.
The 311 hidden diseases were chosen by hand as high-yield topics across NEET PG, INICET, PLAB, USMLE and AMC. These are not official exam lists.