Metagenomics

EMG produced TPA metagenomics assembly of the PRJNA46333 data set (Gene-Environment Interactions at the Skin Surface).

Microbial functional profiles from the Gene-Environment Interactions at the Skin Surface project. The benchmark uses MGnify InterPro abundances from assembled sequences to predict the recorded environmental material associated with each sample.

Binary classificationLicence: PublicMGYS00005379Target: environment (material)mgnify ↗

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)	0.7289	0.7047	0.6778	0.927
LimiX2	0.7168	0.597	0.7695	0.8912
RealTabPFN 2.5	0.6821	0.6347	0.6271	0.9064
Logistic Regression	0.6449	0.5851	0.5859	0.9325
TabFM	0.6212	0.6348	0.5856	0.9225
GeneICL	0.6081	0.4282	0.7758	0.9051
TabDPT	0.6014	0.5429	0.5525	0.9132
Logistic Regression (tuned)	0.6006	0.5025	0.5543	0.8342
AutoGluon	0.5989	0.5377	0.5579	0.9177
TabPFN 3.5	0.5973	0.5275	0.5529	0.9167
MLP	0.5956	0.4592	0.5754	0.9044
RealTabPFN v2	0.5911	0.5604	0.5408	0.9287
MLP (tuned)	0.5718	0.4678	0.5455	0.8327
TabPFN Wide 5k (ne3)	0.5506	0.4486	0.5005	0.904
Causilo	0.55	0.4915	0.5002	0.8662
RealMLP	0.5282	0.4253	0.5289	0.8998
Random Forest (tuned)	0.5009	0.3443	0.4703	0.8046
Extra Trees (tuned)	0.4844	0.3161	0.4588	0.8205
Random Forest	0.4486	0.2692	0.4233	0.8804
XGBoost (tuned)	0.4097	0.1958	0.4038	0.7809
XGBoost	0.404	0.1831	0.3955	0.8177
Extra Trees	0.4016	0.2074	0.3869	0.896
MITRA	0.3953	0.1733	0.3894	0.8078
TabM	0.3283	0.03405	0.341	0.7638