How does a BLAST search work? Put the steps in order.

Attempt every possible global alignment, calculate p $-$values for each alignment, select the alignment with the lowest chance of type $1$

statistical error $($false positive$).$

Split a sequence into words, find improbable local alignments to a database using a scoring matrix, extend words to make HSP$\text{'}$s$,$

calculate e$-$values, discard matches over the threshold.

Add enough gaps to make every nucleotide or amino acid line up with a sequence from the database, iterate the process, construct a

neighbor$-$joining family tree, report results below the e$-$value threshold.

Like google $-$ use several smaller computers or CPU threads to run searches over small datasets, aggregate the results, and rank

results according to a proprietary algorithm.

Spilit a sequence into words, use a scoring matrix to calculate the probability of each word randomly appearing, report the matches

with e$-$values below the threshold.