Bowtie2 --score-min calculation

[kristian_ullrich]

Hi,
I wonder if someone could explain the large differences between the theshold values calculated by the --end-to-end (L,-0.6,-0.6) vesus the --local (G,20,8) min. threshold scores.
I have attached the calculated min. thresholds for both functions (-f(x) = -0.6 + -0.6 * (x) vs. f(x) = 20 + 8.0 * ln(x)).
Best regards
Kristian

[dpryan]

With end-to-end you're requiring that the whole length of the read get aligned, the penalties are then only due to mismatches or indels. With local alignment, you can align just a subset of each read, so it makes sense that the minimum score plateaus fairly quickly. The underlying algorithms for local vs. global alignment are such that you need positive scores for the former and negative (or 0) scores for the latter.

[kristian_ullrich]

Thanks for the quick response and clearifying it. Post closed.