Metagenomics

Human feces metagenome 16s rDNA sequencing

Human fecal microbiome samples from a study of HIV infection, originally described as 16S rRNA sequencing. The benchmark uses MGnify InterPro functional-abundance profiles to predict recorded host sex.

Binary classificationLicence: PublicMGYS00001255Target: host sexmgnify ↗

At this operating point

Model comparison

p=10,000, n=100 · Higher Macro-F1 is better

Random Forest is marked with a diamond. Models are ordered by the selected metric.

Across sample budgets

Learning curves

Models

All metric values at this operating point

Model	Macro-F1	MCC	Balanced accuracy	ROC-AUC
AutoGluon	0.7827	0.6119	0.856	0.9046
GeneICL	0.7259	0.4913	0.7722	0.8742
TabDPT	0.7234	0.4756	0.7217	0.8897
TabPFN 3.5	0.6981	0.4357	0.701	0.8715
TabPFN Wide 5k (ne3)	0.6963	0.4199	0.6932	0.8627
MLP (tuned)	0.6962	0.4031	0.7055	0.8828
LimiX2	0.6962	0.4425	0.761	0.8274
MLP	0.693	0.4291	0.6888	0.8784
TabM	0.6896	0.4019	0.6883	0.8501
RealTabPFN 2.5	0.6876	0.3888	0.6715	0.8729
Kumo Tabular (M)	0.6842	0.386	0.6678	0.8735
RealMLP	0.67	0.3695	0.6717	0.8859
Logistic Regression	0.6479	0.3385	0.6382	0.8964
Random Forest	0.6422	0.3569	0.6255	0.8694
Logistic Regression (tuned)	0.6345	0.3536	0.6298	0.8964
Extra Trees (tuned)	0.6309	0.2896	0.6173	0.8531
Random Forest (tuned)	0.612	0.3148	0.609	0.8508
MITRA	0.5767	0.2191	0.576	0.8317
RealTabPFN v2	0.5726	0.2199	0.5758	0.8913
TabFM	0.5473	0.2079	0.5667	0.8978
XGBoost	0.5463	0.1247	0.5592	0.7967
Causilo	0.5316	0.1649	0.5465	0.8671
Extra Trees	0.5282	0.1707	0.5418	0.8634
XGBoost (tuned)	0.4925	0.05256	0.5212	0.7943