Lab 6

Blast:

Coffee Break

• coffee break

  Using the Coffee Break link above, read the October 22, 2001 entry on RFS. This is supposed to be a good example of how similarity searching can be useful in finding a gene/protein that may be causing or contributing to a genetic disease. Be sure to follow the links HSPC, hypothetical protein, and related sequences. A hypothetical protein is a protein sequence derived from a DNA gene sequence when there is no knowledge that the derived amino acid sequence is actually a protein sequence.
  Now we will attempt to use BLAST to get the same conclusions they report in Coffee break. This is not so easy as it might seem.

  Exercises:

  1. What is the genbank accession number for this hypothetical protein? In the coffee break article, it says that ultimately they figured out that the hypothetical protein was related to Psr2P. What is the genebank accession number for Psr2P? Is it in the list of related sequences for the hypothetical protein? If so, where is it in the list? In NCBI or Entrez documantation, can you find an explanation for how ``related sequences" are found? I could not. Explain what the cd option in BLAST does? Explain what the nr database is?
  2. Copy the amino acid sequence of the hypothetical protein. Remove the base counts on the left and the spaces in the sequence to create a single sequence. Open a BLAST session (details from lab 5) and do a BLAST search with the hypothetical protein sequence. Use Swissprot and also nr for your choice of databases. BLAST may return the same sequence you input as the best match (this therefore will be an exact match). Excluding that one, if it happens, examine the other sequences that BLAST returns. How many are there? What is the range of E-values and scores returned? Is there a relationship between where the local alignment regions are and the scores and E-values? Does such a correlation suggest that BLAST has found a biologically meaningful hit? That is, is the correlation useful information on top of the scores and E-values?
     
     Do the names of the best matching sequences suggest a function for the hypothetical protein? How do the results that BLAST gives compare to what is claimed in the Coffee break article, and what you see in the ``related sequences" page linked from coffee break? Did Blast return the same sequences that Coffee Break reported as related sequences? If not, why not? Did it include the Psr2P sequence, and if so, where in the list was it?
     Now restrict the database (organism) to yeast (Saccharomyces cerevisiae) and do the searches and report the results.
     Now take just the first half of the hypothetical protein and do the search and report the results. Now take just the second half of the hypothetical protein and do the search and report the results. You can do these by specifying the subsequence interval in the BLAST search.
  3. What have you learned about the subtleties of using BLAST from this exercise?
  4. Remember what I discussed in class about using database searchers as a filter to generate clues that you have to check out, versus using database searchers as oracles which give you ``the answer". How does this exercise relate to that discussion?
     If you are really motivated, use the CD results that BLAST returned to learn more about those CD's to see if they may be related to what is known about the biology of Fanconi disease. Try searching the web for a connection between the CD and kindney disease, for example. How does one realistically go from the hypothetical protein to Psr2P via database search?

  The start of a toy BLAST program in Perl:

  Get the four kmer programs:

  Four kmer programs.

  Another kmer program, useful next week
  kmer4.pl

  The new elements of Perl include the substr function, the length function, two-dimensional hashes, the defined function, building up a list as a hash value, and maybe some others. You should read Johnson on these elements of Perl, although I will talk about them explicitly in class. Eventually, you will need to understand all of these for the larger BLAST program we will write. For this lab, kmerfirst.pl will be the most useful. Program kmerfirst.pl finds the first position of each of the different kmers of length k. You will extract what you need from this, and then expand on it, to write the start of a toy BLAST program. Your program should do the following things:

  1. Read in from file a query string Q.

  2. For k = 4, use program kmerfirst.pl to find the first location of each different k-mer in Q. This information is kept in a hash as in program kmerfirst.pl.

  3. Successively read in one string at a time from a file called toygenbank. You should use toygenbank. When you hand in the lab for this week, be sure to show the program working on this toy data, but you can also make up more challenging data to add to it.

  When a string S is read in, scan through it 4-mers, using the same hash as before. For this, extract and adapt what you need from kmerfirst.pl

  4. Whenever (if ever) a 4-mer in S is determined to be in Q, extract the location of the first occurrence of that 4-mer in Q. Then put the characters of Q and S in arrays (as we did in needleman.pl) so that you can examine individual characters. Then scan left from the k-mer in Q and in S, as long as you find matching characters. Repeat to the right. Let L denote the length of the whole match obtained in this way. If L is greater than 10, then print a message that a good HSP has been found between Q and S, and print S.

  Notice that the same HSP gets reported multiple times. Explain why that happens.

  If you want to work ahead, next week the lab will ask you to modify and use kmer3.pl instead of kmerfirst.pl, so that every location of each different 4-mer in Q is collected. Then when a 4-mer common to Q and S is found, the left and right scan is done at each location in Q. Also, the program should use a hash to avoid reporting the same HSP multiple times.

  Here is a suggestion for a different way to repeatedly remove the first character of a string. Try test-it.pl

  ^M