Fig. 5
From: SeqScreen: accurate and sensitive functional screening of pathogenic sequences via ensemble learning
Positive label precision and recall per FunSoC for the four ML models Bl. SVC + NN (OS) (in blue), TS NN (in green), TS Bl. SVC (in yellow), and MV ensemble (in brown). Precision is in solid lines and recall is in dotted lines. TS Bl. SVC shows the best overall recall, whereas TS NN consistently has the highest precision across most of the 32 FunSoCs. In hard-to-classify FunSoCs like nonviral invasion and bacterial counter signaling, TS NN performs poorly indicating a model with a high degree of variance. Similarly, TS Bl. SVC suffers from poor precision in most cases. The majority vote classifier improves on the Bl. SVC + NN (OS) and finds an optimal balance between precision and recall across all FunSoCs