Update pairwise-alignment.md

book/fundamentals/pairwise-alignment.md

@@ -210,7 +210,7 @@ I suggested above that you keep a list of assumptions that are made by this appr
 1. We're scoring all matches as 1 and all mismatches as 0. This suggests that all matches are equally likely, and all mismatches are equally unlikely. What's a more biologically meaningful way to do this (think about protein sequences here)?
 2. Similarly, every gap that is introduced results in the same penalty being incurred. Based on what we know about how insertion/deletion events occur, what do you think is a more biologically meaningful way to do this?
 
-All scoring schemes have limitations, and you should remember that when you're working with software that generates alignments for you (e.g., systems such as [BLAST](http://blast.ncbi.nlm.nih.gov/Blast.cgi)). Especially as you're getting started in bioinformatics, it's easy to forget that and just accept the result from computer software as "the right answer". You'll need to determine if you agree with the result that a computational system gives you, which will involve examining the result in the context of what you know about the biology of the systems your studying. Algorithms such as the one we just explored are there to help you do your work, but they won't do your work for you. Their answers are based on models (for example, how we model matches, mismatches and gaps here) and as you're learning here, the models are not perfect. Be skeptical!
+All scoring schemes have limitations, and you should remember that when you're working with software that generates alignments for you (e.g., systems such as [BLAST](http://blast.ncbi.nlm.nih.gov/Blast.cgi)). Especially as you're getting started in bioinformatics, it's easy to forget that and just accept the result from computer software as "the right answer". You'll need to determine if you agree with the result that a computational system gives you, which will involve examining the result in the context of what you know about the biology of the systems you're studying. Algorithms such as the one we just explored are there to help you do your work, but they won't do your work for you. Their answers are based on models (for example, how we model matches, mismatches and gaps here) and as you're learning here, the models are not perfect. Be skeptical!
 
 Another important consideration as we think about algorithms for aligning pairs of sequences is how long an algorithm will take to run as a function of the input it's provided (or in technical terminology, the [computational complexity](http://bigocheatsheet.com/) of the algorithm). When searching a sequence against a database (for example, to get an idea of what its function is), you may have billions of bases to search against, which would correspond to billions of columns in one of the matrices we just computed. Computers are fast, but the data sets you're going to be working with are very large and in many cases growing exponentially in size over time. Working in bioinformatics, it's inevitable that you're going to begin to discover the limitations of the algorithms and software you use. Runtime and memory requirements are the usual culprits. Because the data sets are getting bigger more quickly than computers are getting faster (at least as of this writing), just waiting for computers to get faster won't work. We need smart people who understand some computer science and some biology to design clever algorithms, software, and analytic techniques to enable the next generation of advances that technologies like high-throughput DNA sequencing are promising. (And there are a lot of people who want to spend good money to pay people who can do these things, so keep reading!)