Skip to main content
. 2020 Apr 13;36(12):3637–3644. doi: 10.1093/bioinformatics/btaa242

Fig. 1.

Fig. 1.

Balanced accuracy for models at different COSMIC recurrence levels shows that the coding models achieve a peak validation accuracy of 74.1% at a recurrence threshold of ρ = 7 (top), whereas the non-coding models achieve a peak accuracy of 69.5% at a recurrence threshold of ρ = 8 (bottom)