Metagenomics

Human skin bacterial metagenome and Virome Metagenome

Skin microbiome data from a study that sequenced both virus-like particles and total microbial communities. The benchmark uses MGnify InterPro functional-abundance profiles to predict the recorded sex of the human host.

Binary classificationLicence: PublicMGYS00005384Target: 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
Kumo Tabular (M)	0.5348	0.08049	0.541	0.5542
Extra Trees (tuned)	0.5304	0.08047	0.5401	0.5257
TabPFN 3.5	0.5264	0.07015	0.5355	0.5477
RealTabPFN 2.5	0.5244	0.05948	0.5306	0.5367
MLP	0.5109	0.03765	0.5181	0.5303
Extra Trees	0.5055	0.03181	0.5157	0.5178
TabFM	0.5052	0.04751	0.5233	0.5197
XGBoost (tuned)	0.5026	0.02202	0.5109	0.5219
Random Forest	0.501	0.01297	0.5063	0.5116
RealTabPFN v2	0.5008	0.03659	0.5182	0.5299
Logistic Regression	0.4924	0.00307	0.502	0.4855
Random Forest (tuned)	0.4888	-0.01064	0.4944	0.5038
XGBoost	0.4863	-0.003825	0.4985	0.486
GeneICL	0.4788	0.02677	0.5099	0.5104
MLP (tuned)	0.4785	0.04325	0.5129	0.5093
TabM	0.4777	-0.01356	0.4941	0.4987
Logistic Regression (tuned)	0.4614	-0.05052	0.4757	0.4795
MITRA	0.4554	-0.006662	0.5061	0.4966
TabDPT	0.4504	-0.08234	0.4627	0.4587
Causilo	0.4326	-0.04227	0.4856	0.4964
TabPFN Wide 5k (ne3)	0.4288	-0.02044	0.4917	0.5124
AutoGluon	0.4285	-0.03993	0.4845	0.4542
LimiX2	0.4167	-0.07122	0.4814	0.4444
RealMLP	0.3857	0.04258	0.5224	0.5309