Non-parametric and semi-parametric support estimation using SEquential RESampling random walks on biomolecular sequences

Table 6 Support estimation method performance on long-gap-length model conditions

Model condition	PR-AUC (%)
Model condition	GUIDANCE2	SERES + GUIDANCE2	Pairwise t-test corrected q-value
10.long.A	92.32	92.94	\(9.7 \times 10^{-4}\)
10.long.B	90.62	91.64	\(3.3 \times 10^{-6}\)
10.long.C	85.10	87.93	\(9.7 \times 10^{-4}\)
10.long.D	79.22	86.18	\(9.7 \times 10^{-4}\)
10.long.E	67.63	78.48	\(9.7 \times 10^{-4}\)

Model condition	ROC-AUC (%)
Model condition	GUIDANCE2	SERES + GUIDANCE2	DeLong et al. test corrected q-value
10.long.A	89.99	90.99	\(<10^{-10}\)
10.long.B	91.84	93.02	\(<10^{-10}\)
10.long.C	93.14	94.59	\(<10^{-10}\)
10.long.D	93.89	96.13	\(<10^{-10}\)
10.long.E	92.62	94.38	\(<10^{-10}\)

The performance of GUIDANCE2 and SERES + GUIDANCE2 is compared across model conditions 10.long.A through 10.long.E (named in order of generally increasing sequence divergence). Aggregate PR-AUC and ROC-AUC are reported across all replicate datasets in a model condition (\(n=20\)), and the best AUC for each pairwise method comparison on a model condition is shown in italics. Statistical significance of PR-AUC or ROC-AUC differences was assessed using a one-tailed pairwise t-test or DeLong et al. [5] test, respectively, and multiple test correction was performed using the method of Benjamini and Hochberg [1]. Corrected q-values are reported (\(n=20\)) and all were significant (\(\alpha =0.05\))