Statistics

Statistical Insights

Key Insights and Analytics from the Application Backend .

Phase I — Housekeeping vs Non-housekeeping

Phase I decides whether a gene is Housekeeping (HK) or Non-housekeeping (NHK). SVM (GridSearchCV Best) is the headline model.

Model	Train accuracy	Test accuracy	Precision	Recall	F1	5-fold CV
SVM (Voting, 4 kernels)	0.9498	0.9459	0.9459	0.9459	0.9459	0.9491 ± 0.0010
SVM (GridSearchCV Best)	0.9778	0.9768	0.9768	0.9768	0.9768	0.9741
Random Forest	0.8400	0.8386	0.8749	0.8386	0.8337	0.8379 ± 0.0149
XGBoost	0.8370	0.8356	0.8607	0.8356	0.8319	0.8369 ± 0.0138
Elastic Net	0.7150	0.7131	0.7263	0.7131	0.7103	0.7442 ± 0.0194

Phase II — NHK tissue assignment (One-vs-Rest)

Phase II runs only for NHK genes and assigns one or more tissue groups. Macro-F1 highlights performance across classes; micro-F1 reflects aggregate label decisions.

Model	Precision macro	Recall macro	F1 macro	F1 micro	CV F1 macro	CV F1 micro
SVM (OVR)	0.9133	0.7564	0.8272	0.8349	0.8304	0.8401
Random Forest (OVR)	0.9681	0.5699	0.7103	0.7347	0.7201	0.7438
XGBoost (OVR)	0.9537	0.7377	0.8299	0.8408	0.8257	0.8393
Elastic Net (OVR)	0.8295	0.3716	0.4412	0.4808	0.4166	0.4245

TSI Distribution & Expression Heatmap

Tissue Specificity Index (TSI) Distribution

Expression Data Heatmap

Functional Annotation & WGCNA

Functional Annotation — Flower Parts

Functional Annotation — Root Tissues

Functional Annotation — Seed Tissues

WGCNA Heatmaps

Functional Annotation — Flower Development Stages

Functional Annotation — Green Tissues