Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
SEEING CONSERVED SIGNALS: USING ALGORITHMS TO DETECT SIMILARITIES BETWEEN BIOSEQUENCES 81 memory that can be loaded with the scores for δ(ai, ?), δ (â, ?), and δ (ai, â) where ? is any symbol in the underlying alphabet Ï. The beauty of the systolic array is that it can perform comparisons of A against a stream of B sequences, processing each symbol of the target sequences in constant time per symbol. With current technology, chips of this kind operate at rates of 3 million to 4 million symbols per second. A systolic array of 1,000 of these simple processors computes an aggregate of 3 billion to 4 billion dynamic programming entries per second. COMPARING ONE SEQUENCE AGAINST A DATABASE The current GENBANK database (Benson et al., 1993) of DNA sequences contains approximately 191 million nucleotides of sequence in about 183,000 sequence entries, and the PIR database (Barker et al., 1993) of protein sequences contains about 21 million amino acids of data in about 71,000 protein entries. Whenever a new DNA or protein sequence is produced in a laboratory, it is now routine practice to search these databases to see if the new sequence shares any similarities with existing entries. In the event that the new sequence is of unknown function, an interesting global or local similarity to an already-studied sequence may suggest possible functions. Thousands of such searches are performed every day. In the case of protein databases, each entry is for a protein between 100 and 1,500 amino acids long, the average length being about 300. The entries in DNA databases have tended to be for segments of an organism's DNA that are of interest, such as stretches that code for proteins. These segments vary in length from 100 to 10,000 nucleotides. The limited length here is not intrinsic to the object as in the case of proteins, but because of limitations in the technology and the cost of obtaining long DNA sequences. In the early 1980s the longest consecutive stretches being sequenced were up to 5,000 nucleotides long. Today the sequences of some viruses of length 50,000 to 100,000 have been determined. Ultimately, what we will have is the entire sequence of DNA in a chromosome (100 million to 10 billion nucleotides), and entries in the database will simply be annotations describing interesting parts of these massive sequences.