Figure 2

Regression analysis output (slope $\widehat{\mathit{\beta }}$ and coefficient of determination r²) for both SMART (version 6) and Pfam (release 27) domains. Figure A and B depict the histograms of the slopes $\widehat{\mathit{\beta }}$ for the original versus reconstructed scores for SMART domains calculated for HMMER2 and HMMER3 respectively while Figure C and D depict the histograms of the slopes $\widehat{\mathit{\beta }}$ for the Pfam domains. The HMMER2 results exhibit high reproducibility at an average $\widehat{\mathit{\beta }}$ of 1.000 ± 0.001 (SMART) and 1.000 ± 0.002 (Pfam) while HMMER3 results also show good, though slightly worse reproducibility with average $\widehat{\mathit{\beta }}$ of 1.015 ± 0.017 (SMART) and 1.017 ± 0.013 (Pfam). Figures E, F, G and H shows the corresponding histograms for the goodness of fit, in terms of r². Similarly, the HMMER2 reconstruction exhibits excellent fit at an average r² of 1.000 ± 0.003 (SMART) and 1.000 ± 0.007 (Pfam). HMMER3 reconstruction closely followed at an average r² of 0.997 (SMART) and 0.998 (Pfam) over a slightly larger variation of 0.007 (SMART/Pfam). In hindsight, all values of $\widehat{\mathit{\beta }}$ and r² converges to one with little variation and this implies that the reconstruction workflow for HMMER2/3 scores are highly reproducible.