Biol425 2015

Computational Molecular Biology (BIOL 425/790.49, Spring 2015) Instructor: Dr Weigang Qiu, Associate Professor of Biology; weigang at genectr.hunter.cuny.edu Room:1000G HN (10th Floor, North Building, Computer Science Department, Linux Lab FAQ) Class Hours: Wednesdays 10:10 am-12:40 pm Office Hours & Address: Wednesdays 5-7pm or by appointment; Address: Belfer Research Building, 413 East 69th Street (between York and 1st Avenue), New York, NY 10021. Direction via Google Map;

---

Course Schedule (All Wednesdays)

January 28. Course overview & Unix tools

• Course Overview. Lecture slides:
  File:Session-1-small-2015.pdf
• Learning Goals: (1) Understand the "Omics" files; (2) Review/Learn Unix tools
• In-Class Tutorial: Unix file filters

1. Without changing directory, long-list genome, transcriptome, and proteome files using ls
2. Without changing directory, view genome, transcriptome, and proteome files using less -S
3. Using grep, find out the size of Borrelia burgdorferi B31 genome in terms of the number of replicons ("GBB.1con" file)
4. Using sort, sort the replicons by (a) contig id (3rd field, numerically); and (b) replicon type (4th field)
5. Using cut, show only the (a) contig id (3rd field); (b) replicon type (4th field)
6. Using tr, (a) replace "_" with "|"; (b) remove ">"
7. Using sed, (a) replace "Borrelia" with abbreviation "B."; (b) remove "plasmid" from all lines
8. Using paste -s, extract all contig ids and concatenate them with ";"
9. Using a combination of cut and uniq, count how many circular plasmids ("cp") and how many linear plasmids ("lp")

• In-Class Challenges

1. Using the "GBB.seq" file, find out the number of genes on each plasmid

grep ">" ../../bio425/data/GBB.seq | cut -c1-4| sort | uniq -c

1. Using the "ge.dat" file, find out (a) the number of genes; (b) the number of cell lines; (c) the expression values of three genes: ERBB2, ESR1, and PGR

grep -v "Description" ../../bio425/data/ge.dat | wc -l; or:  grep -vc "Description" ../../bio425/data/ge.dat
grep "Description" ../../bio425/data/ge.dat | tr '\t' '\n'| grep -v "Desc" | wc -l
grep -Pw "ERBB2|PGR|ESR1" ../../bio425/data/ge.dat

Assignment #1

Unix Text Filters (10 pts) Show both commands and outputs for the following questions: Without changing directory (i.e., remain in your home directory), locate and long-list the genbank file named "GBB.gb" in the course data directory Count the total number of lines, show the first and last 10 lines of the file. Using a combination of head and tail commands, show only the lines containing the translated protein sequence of the first gene Count the total number of replicans by extracting lines containing "LOCUS" (case sensitive); sort them by the total number of bases ("bp") Remove the string "(plasmid" from the above output Extract the second column (replicon names) from the above output. [Hint: these fields are delimited by an unequal number of spaces, not by tabs. Use tr -s to first squeeze to single space]

---

February 4. Genomics (1): Gene-Finding

• Learning goals: (1) Running UNIX programs; (2) Parse text with Perl anonymous hash
• Lecture slides:
  File:Session-2-2015.pdf
• In-Class Tutorials

1. Identify ORFs in a prokaryote genome
   1. Go to NCBI ORF Finder page
   2. Paste in the GenBank Accession: AE000791.1 and click "orfFind"
   3. Change minimum length for ORFs to "300" and click "Redraw". How many genes are predicted? What is the reading frame for each ORF? Coordinates? Coding direction?
   4. Click "Six Frames" to show positions of stop codons (magenta) and start codons (cyan)
2. Gene finder using GLIMMER
   1. Locate the GLIMMER executables: ls /data/biocs/b/bio425/bin/
   2. Locate Borrelia genome files: ls /data/biocs/b/bio425/data/GBB.1con-splitted/
   3. Predict ORFs: ../../bio425/bin/long-orfs ../../bio425/data/GBB.1con-splitted/Borrelia_burgdorferi_4041_cp9_plasmid_C.fas cp9.coord [Note the two arguments: one input file and the other output filename]
   4. Open output file with cat cp9.coord. Compare results with those from NCBI ORF Finder.
   5. Extract lines with coordinates" grep "^[0-9]" cp9.coord > cp9.coord2
   6. Extract sequences into a FASTA file: ../../bio425/bin/extract ../../bio425/data/GBB.1con-splitted/Borrelia_burgdorferi_4041_cp9_plasmid_C.fas cp9.coord2 > cp9.fas [Note two input files and standard output, which is then redirected (i.e., saved) into a new file]
3. bioseq
   1. Add bp-utils into $PATH by editing the .bash_profile file and run source .bash_profile
   2. Run bioseq with these options: -l, -n, -c, -r, -t1, -t, -t6, and -s
   3. In-Class Challenge: use bioseq to extract and translate the 1st (+1 frame) and 5th (-1 frame) genes.

cp ../../bio425/data/GBB.1con-splitted/Borrelia_burgdorferi_4041_cp9_plasmid_C.fas cp9.plasmid.fas # copy plasmid sequence
bioseq -s'163,1272' cp9.plasmid.fas | bioseq -t1 # extract first predicted ORF (+1 frame)
bioseq -s'3853,4377' cp9.plasmid.fas | bioseq -r | bioseq -t1 # extract 5th ORF (since it is -1 frame, need to rev-com before translation

Assignment #2 (Finalized as of Saturday 2/7/205, at 1pm)

UNIX Exercises (10 pts) Using ssh, log into eniac.cs.hunter.cuny.edu & then ssh into another host (e.g. "cslab8") for the exercises Copy data file ../../bio425/data/mystery_seq1.fas into your home directry Add /data/biocs/b/bio425/bin to your the $PATH variable in the ".bash_profile" file (using vi or emacs). Run source ~/.bash_profile to activate the above path Use the long-orf command to predict ORFs on the mystery_seq1. Save resulting coord file as "mystery_seq1.coord". Extract ONLY the lines containing ORF information into a new file "mystery_seq1.coord2" using grep Run extract to extract FASTA sequences of all predicted ORFs. Save as "mys.nuc". Translate "mys.nuc" with bioseq -t1. Save to "mys.pep" (don't be surprised if the 2nd sequence contains internal stop codons) Use bioseq to extract the ORF sequences, save as "mys.nuc2" Translate "mys.nuc2" with bioseq -t1. Save to "mys.pep2". This file should NOT contain any internal stop codons. Note: Show all commands and outputs.To log off, type "exit" twice (first log out of the "cslab" host, 2nd time log off the "eniac" host).

---

February 11. Genomics (2): BASH & BLAST

• Lecture Slides:
  File:Session-3-small-2015.pdf
• Learning goal: (1) BASH scripting; (2) Homology searching with NCBI blast
• In-Class exercises

1. Build workflow with BASH scripts

#!/usr/bin/bash
bioseq -s"163,1272" cp9.plasmid.fas > gene-0001.nuc;
bioseq -t1 gene-0001.nuc > gene-0001.pep;
exit;

Save the above as "extract-gene-v1.bash". Make it executable with chomd +x extract-gene-v1.bash. Run it with ./extract-gene-v1.bash

1. Improvement 1. make it work for other genes (by using variables)

#!/usr/bin/bash
id=$1;
begin=$2;
end=$3;
bioseq -s"$begin,$end" cp9.plasmid.fas > gene-$id.nuc;
bioseq -t1 gene-$1.nuc > gene-$id.pep;
exit;

Save the above as "extract-gene-v2.bash". Make it executable with chomd +x extract-gene-v2.bash. Run it with ./extract-gene-v2.bash 0001 163 1272 (Note three arguments). Question: How to modify the code to make it work for genes encoded on the reverse strand?

1. Improvement 2. make it work for both positive and negative genes (by using an "if" statement)

#!/usr/bin/bash
id=$1;
begin=$2;
end=$3;
strand=$4;
bioseq -s"$begin,$end" cp9.plasmid.fas > gene-$id.nuc;
if [ $strand -gt 0 ]; then
    bioseq -t1 gene-$1.nuc > gene-$id.pep;
else
    bioseq -r gene-$1.nuc > gene-$id.rev;
    bioseq -t1 gene-$1.rev > gene-$id.pep;
fi;
exit;

1. Improvement 3. Make it read "cp9.coord" as an input file, by using a "while" loop:

#!/usr/bin/bash
cat $1 | grep "^[0-9]" | while read line; do
    id=$(echo $line | cut -f1 -d' '); # run commands and save result to a variable
    x1=$(echo $line | cut -f2 -d' ');
    x2=$(echo $line | cut -f3 -d' ');
    strand=$(echo $line | cut -f4 -d' ');
    if [ $strand -lt 0 ]; then
        bioseq -s"$x2,$x1" cp9.plasmid.fas > gene-$id.nuc;
        bioseq -r gene-$id.nuc > gene-$id.rev;
        bioseq -t1 gene-$id.rev;
    else
        bioseq -s"$x1,$x2" cp9.plasmid.fas > gene-$id.nuc;
        bioseq -t1 gene-$id.nuc;
    fi;
done;
exit;

1. Run Stand-alone BLAST:

• Add blast binaries to your $PATH in ".bash_profile": export PATH=$PATH:/data/biocs/b/bio425/ncbi-blast-2.2.30+/bin/
• BLAST tutorial 1. A single unknown sequence against a reference genome

cp ../../bio425/data/GBB.pep ~/. # Copy the reference genome
cp ../../bio425/data/unknown.pep ~/. # Copy the query sequence
makeblastdb -in GBB.pep -dbtype prot -parse_seqids -out ref # make a database of reference genome
blastp -query unknown.pep -db ref # Run simple protein blast
blastp -query unknown.pep -db ref -evalue 1e-5 # filter by E values
blastp -query unknown.pep -db ref -evalue 1e-5 -outfmt 6 # concise output
blastp -query unknown.pep -db ref -evalue 1e-5 -outfmt 6 | cut -f2 > homologs-in-ref.txt # save a list of homologs
blastdbcmd -db ref -entry_batch homologs-in-ref.txt > homologs.pep # extract homolog sequences

• In-Class challenges

Assignment #3 (Finalized on Feb 15, 2015, 2AM)

BASH Exercises (20 pts) Note: Log on your eniac account and then ssh into a "cslab" host (e.g., ssh cslab5). Perform the following tasks without changing directory (i.e., stay in your home directory). UNIX filter exercises (5 pts) For the codon file /data/biocs/b/bio425/data/codon.table.T, sort by single-letter code of amino acids use "grep" to show only codons that start with an "A" Show all codons for "Arg", without amino acid symbols Use an editor, compose a BASH script (name it "read-codon.bash") that uses a "while" loop to read each line of this file. First, use "grep -v" to exclude stop codons. Then, create three variables to capture the values of three elements on each line. For each line, print the output in the following format: "AAA codes for Lys (K)". Exercise for the use of wildcard & BASH "for" loop (5 pts) Long-list (i.e., ls -l) the following file in the "/data/biocs/b/bio425/data/GBB.1con-splitted/" directory: Borrelia_burgdorferi_3615_main.fas (this file contains the main chromosome sequence of the B31 strain of Lyme disease bacteria). Run bioseq -l to find out the length of this sequence. Without changing directory, long-list ALL files with the ".fas" suffix in the same directory (using wildcard). This command list all replicons (chromosome and 21 plasminds) of this genome. Write a BASH "for" loop on the command line (do not use an editor) to obtain length of each replicon using bioseq -l Exercise for BASH script (5 pts): Compose a pipeline of predicting and extracting gene sequences using a single BASH script. (You may name this script "predict-orfs.bash"). The script takes a genome sequence file as a single input. It outputs predicted ORF sequences in two files: "orf.nuc" & "orf.pep". Copy ../../bio425/data/GBB.1con-splitted/Borrelia_burgdorferi_4091_cp26_plasmid_B.fas file to your home directory and predict all ORFs on this plasmid. Note your resulting "orf.pep" file should not contain any stop codons except as the last. BLAST exercise (5 pts). Use "blastp" to identify matches of the sequence ../../bio425/data/unknown.pep in the "ref" database (which was created in class), using an e-value cutoff of 1e-10 Show all pairwise alignments Show output in a tabular format (using "-outfmt 6") Explain the following blast terms: "Identity", "Positives", and "Expect". [Hint, visit the NCBI BLAST home page and this FAQ].

February 25. Genomics (3). Homology searching with BLAST

• Lecture Slides
• Learning goals: Homology, BLAST, & Alignments
• In-Class Challenge. Compose a BioPerl-based script ("translation.pl") for translating a DNA sequence. Input: a DNA file in FASTA (e.g., "bio425/data/TenSeq.nuc") . Output: protein sequences in FASTA. Once the code is working, add a regular expression to skip and warn users for any translated sequence that doesn't start with with a start codon ("ATG" or "TTG"), or doesn't end with a stop codon ("TAG", "TAA", "TGA"), or contains internal stop codons.
• BLAST tutorial 2. Find homologs within the new genome itself

cp ../../bio425/data/N40.pep ~/. # Copy the unknown genome
makeblastdb -in N40.pep -dbtype prot -parse_seqids -out N40 # make a database of the new genome
blastp -query unknown.pep -db N40 -evalue 1e-5 -outfmt 6 | cut -f2 > homologs-in-N40.txt # find homologs in the new genome
blastdbcmd -db N40 -entry_batch homologs-in-N40.txt >> homologs.pep # append to homolog sequences

• BLAST tutorial 3. Multiple alignment & build a phylogeny

../../bio425/bin/muscle -in homologs.pep -out homologs.aln # align sequences
cat homologs.aln | tr ':' '_' > homologs2.aln
../../bio425/bin/FastTree homologs2.aln > homologs.tree # build a gene tree
../../bio425/figtree &  # view tree (works only in the lab; install your own copy if working remotely)

• BLAST tutorial 4. Annotate the entire genome

blastp -query N40.pep -db ref -evalue 1e-5 -outfmt 6 > blast.fwd # foward blast
blastdbcmd -db ref -entry all > ref.pep
blastp -query ref.pep -db N40 -evalue 1e-5 -outfmt 6 > blast.rev # reverse blast

Assignment #4 (Finalized Sat 3/1)
BLAST exercise (5 pts) Run BLASTp to identify all homologs of BBA18 in the reference genome ("ref"). First, obtain BBA18 peptide sequence using blastdbcmd -db ref -entry 'BBA18'. Then, run blastp against the "ref" database with the following e-value cutoffs: 1e-5, 1e-3, and 1e-1. Submit the results for 1e-3 with the "-outfmt 6" option. [Note: If you work remotely, you have to first ssh into "enaic.cs.hunter.cuny.edu" and then ssh again into "cslab1" or other workstations to access the blast programs.]
BioPerl exercise (5 pts) Write a BioPerl-based script ("pick-by-id.pl") to extract sequences by id. It will emulate the "blastdbcmd -db -entry" command. Use "GBB.pep" and "BBA15" as the two arguments. Usage: ./pick-by-id.pl GBB.pep BBA15. Your script should handle exceptions gracefully, e.g., by printing a warning message if the sequence is not found. #!/usr/bin/perl -w use strict; use lib '/data/biocs/b/bio425/bioperl-live'; use Bio::SeqIO; die "Usage: $0 <a FASTA file> <id>\n" unless @ARGV == 2; my ($file, $id) = @ARGV; my $in = Bio::SeqIO->new(-file=>$file, -format=>'fasta'); my $found = 0; # used to record if sequence is found or not (boolean) while(my $seqobj=$in->next_seq()) { my $seq_id = $seqobj->display_id(); if ($seq_id eq $id) { print ">$seq_id\n", $seqobj->seq(), "\n"; $found = 1; } } warn "No sequence with ID $id is found\n" unless $found; exit;
Read & Respond (5 pts) Read pg 117 (last paragraph)-118 (1st paragraph) & define "ortholog" and "paralog". Draw Figure 2.21(B) by hand and indicate (on the graph) a pair of ortholog and a pair of paralog. Read Box 2.4 (pg 120), introduction & the section labeled as "General principles". Define "homology", "phylogeny", and "branch length".

March 4: Genomics (4). Molecular phylogenetics

• Lecture Slides:
  File:Session-5-small.pdf
• Learning goal: how to interpret, build, and test phylogeny
• Perl challenge: Write a BioPerl script ("get-seq-from-gb.pl") to retrieve a sequence from the GenBank. Follow the example in Bio::DB::GenBank. Input: a GenBank accession (e.g., "J00522"). Output: a file in "genbank" format. Note: include this statement: use lib '/data/biocs/b/bio425/perl5';

#!/usr/bin/perl -w
use strict;
use lib '/data/biocs/b/bio425/bioperl-live';
use lib '/data/biocs/b/bio425/perl5';
use Bio::DB::GenBank;
use Bio::SeqIO;
die "Usage: $0 <genbank_accession, e.g., J00522>\n" unless @ARGV == 1;
my $acc = shift @ARGV;
my $gb = Bio::DB::GenBank->new();
my $seq = $gb->get_Seq_by_acc($acc);
my $out = Bio::SeqIO->new(-file=>">" . $acc.".gb", -format=>'genbank');
$out->write_seq($seq);
exit;

• Tutorials

1. Tree Quizzes:
   File:Baum etal05 Quiz.pdf

   Tree Quizzes
2. Maximum Parsimony: Exercise 2.4 (pg. 123)

Assignment #5 (Finalized, 3/7 Friday 9pm)
Phylogeny questions (5 pts) Referring to the tree at right, answer the following questions. Explain your answer. Which of the five marks corresponds to the most recent common ancestor of the mushroom and the sponge? [d] Is Mouse more related to Sponge or to Mushroom? [Mouse. MRCA of Sponge and Mouse is at e, which is more recent than d, the MRCA of Mouse and Mushroom] Is Mouse more related to Ferm or to Tomato? [Equally related, since MRAC(Mouse-Fern) is at b, so is MRAC(Mouse-Tomato)] Indicate on the tree graph a "clade" and a pair of "sister taxa" [e.g., Clade & Sister: Sponge-Mouse]
BLAST and phylogeny exercise (5 pts) Identify all homologs of BBA68 in the ref and N40 genomes using BLASTp. Construct a phylogeny of BBA68 homologs using MUSCLE and FastTree. View and edit the tree using FigTree (download and install your own copy if you are not using the lab computers). Root the tree on the mid-point by following the right-side panel "Trees"->"Root tree"->"midpoint". Based on the phylogeny, identify N40 orthologs of the following B31 genes: BBA68, BBA69, BBA71, BBA72, and BBA73. (Bonus +2 pts) What are the two possible evolutionary mechanisms that there is no ortholog of BBA70 in the N40 genome? Your answer should consist of the following parts: All BLAST commands A tree rooted on the mid-point as exported by FigTree Labels Indicating each ortholog on the tree (+2 pts) Answer to the bonus question above

March 11: Putting it together: Annotating a new genome

• Review:
  File:Session-6-small.pdf

  Review slides
• Annotate a genome in class: Claim your genome
• No homework assignments this week. Review ALL of your previous assignments

March 18. Midterm Practicum

10:10-12 (No computer access)

March 25 (No Class; Do Assignment #6)

Assignment #6 (Finalized on Friday, 3/21)
Perl/BioPerl Exercises (5+5 pts) The following Perl code dumps nucleotide and codon frequencies given a coding sequence. It is based on the Bio::Tools::SeqStats module of BioPerl. Copy and run it with ./seq-stats.pl ../../bio425/data/B31_ospA.fas to understand its behavior. #!/usr/bin/perl -w use strict; use lib '/data/biocs/b/bio425/bioperl-live'; use Bio::Tools::SeqStats; use Bio::SeqIO; use Data::Dumper; die "Usage: $0 <fasta>\n" unless @ARGV == 1; my $filename = shift @ARGV; my $in = Bio::SeqIO->new(-file=>$filename, -format=>'fasta'); my $seqobj = $in->next_seq(); my $seq_stats = Bio::Tools::SeqStats->new($seqobj); my $monomers = $seq_stats->count_monomers(); my $codons = $seq_stats->count_codons(); print Dumper($monomers, $codons); exit; Write a Perl code ("my-seq-stats.pl") to emulate its two methods ("count_monomers" & "count_codons") with the following specific requirements: Read the fasta file with Bio::SeqIO Write a subroutine named as "count_monomers" to count and print the frequency of each nucleotide (in percentage), sorted by nucleotide type. The argument for the subroutine would be the sequence as a string, as in count_monomers($seqobj->seq). The return value would be a reference to a hash (the same as the $monomers in the above code). Hint: use hash to count unique bases, see Perl Cookbook example, especially the code listed under "4.6.2.4. Faster but different" In a similar way, write a subroutine named as "count_codons" to count and print the frequency of each codon, sorted by codons. Your codon counts should be the same as those produced by the listed code. (Bonus 5 pts, Warning: This may cost you many days of frustration. Don't attempt it if you have other priorities. Finish the "Read and Response" first. For the adventurous type, email me if you are stuck. But I will give partial credits if you do try.) Extend the above code ("seq-stats.pl") to be able to process multiple coding sequences in the file ../../bio425/data/GBB.seq Use the Bio::Tools::SeqStats method "count_codons()"; Skip sequences containing non-ATCG nucleotides (otherwise, you will see long error messages) Dumper total counts for all 64 codons (Expected answer for the first four codons: 'AAA' => 37110, 'AAC' => 6440, 'AAG' => 9875, 'AAT' => 26016) Count and print GC% at all 1st codon positions (Expected answer: 0.3753) Count and print GC% at all 2nd codon positions (Expected answer: 0.2772) Count and print GC% at all 3rd codon positions (Expected answer: 0.2147)
Read & Respond (5 pts) Gene Ontology (Box 2.5, pg 126-127). Describe what is "ontology" and what is "Gene Ontology". For the supplied example of the "Gap1" gene, list a GO term for each of the three GO categories (BP, MF, and CC). Example: Biological Process (BP): "mitosis".

April 1: Transcriptome (1): Univariate statistics

• Learning goals: 1. Microarray technology; 2. Introduction to R
• Lecture Slides:
  File:Session-7-small.pdf
• R Tutorial 1

1. Set a working directory
   1. Using a Linux terminal, make a directory called R-work
   2. Start R: Type "R" on terminal prompt and hit return (Note: This link works only if you use computers in the 1000G lab. Install and use your own R & R Studio for homework assignments)
   3. Find working directory with getwd()
   4. Change working directory with setwd("~/R-work")
2. The basic R data structure: Vector
   1. Construct a character vector with c.vect=c("red", "blue", "green", "orange", "yellow", "cyan")
   2. Construct a numerical vector with n.vect=c(2, 3, 6, 10, 1, 0, 1, 1, 10, 30)
   3. Construct vectors using functions
      1. n.seq1=1:20
      2. n.seq2=seq(from=1, to=20, by=2)
      3. n.rep1=rep(1:4, times=2)
      4. n.rep2=rep(1:4, times=2, each=2)
   4. Use built-in help of R functions: ?seq or help(seq)
   5. Find out data class using the class function: class(c.vect)
3. Access vector elements
   1. A single value: n.vect[2]
   2. A subset: n.vect[3:6]
   3. An arbitrary subset: n.vect[c(3,7)]
   4. Exclusion: n.vect[-8]
   5. Conditional: n.vect[n.vect>5]
4. Apply functions
   1. Length: length(n.vect)
   2. Range: range(n.vect); min(n.vect); max(n.vect)
   3. Descriptive statistics: sum(n.vect); var(n.vect); sd(n.vect)
   4. Sort: order(n.vect). How would you get an ordered vector of n.vect? [Hint: use ?order to find the return values]
   5. Arithmetics: n.vect +10; n.vect *10; n.vect / 10
5. Quit an R session
   1. Click on the "History" tab to review all your commands
   2. Save your commands into a file by opening a new "R Script" and save it as vector.R
   3. Save all your data to a default file .RData and all your commands to a default file ".Rhistory" with save.image()
   4. Quit R-Studio with q()

• R tutorial 2

1. Start a new R studio session and set working directory to ~/R-work
2. Generate a vector of 100 normal deviates using x.ran=rnorm(100)
3. Visualize the data distribution using hist(x.ran)
4. Explore help file using ?help; example(hist)
5. How to break into smaller bins? How to add color? How to change x- and y-axis labels? Change the main title?
6. Test if the data points are normally distributed.[Hints: use qqnorm() and qqline()]
7. Save a copy of your plot using "Export"->"Save Plot as PDF"

• R tutorial 3

1. Using a Linux terminal, make a copy of breast-cancer transcriptome file /data/biocs/b/bio425/data/ge.dat in your R working directory
2. Start a new R studio session and set working directory to ~/R-work
3. Read the transcriptome file using ge=read.table("ge.dat", header=T, row.names=1, sep="\t")
   1. What is the data class of ge?
   2. Access data frame. What do the following commands return? ge[1,]; ge[1:5,]; ge[,1]; ge[,1:5]; ge[1:5, 1:5]
   3. Apply functions: head(ge); tail(ge); dim(ge)
4. Save a copy of an object: ge.df=ge.
5. Transform the transcriptome into a numerical matrix for subsequent operations: ge=as.matrix(ge)
6. Test if the expression values are normally distributed.
7. Save and quit the R session

• In class challenge: Replicate normalization in Figure 4.8 (pg 208)

Assignment #7 (Finalized on Sat, 4/5)

(2 pts) Install R (choose a binary distribution) first and then R Studio to your home/laptop computer [No need to show any output. I trust you on this.] (4 pts) Make 2 histograms based on the "ge" data set: (a) expression values for a single (any one if fine) selected cell line (cell lines are in columns, using small-sized bins, e.g., "br=100") and (b) expression values for a selected (any one if fine) gene (genes are in rows). Show all R commands and the final plots (label the plot title with cell line or gene name) [Notes: You need to download a copy of the ge.dat onto your own computer. On Mac, you could run this command: scp your_user_name@eniac.cs.hunter.cuny.edu:ge.dat ge.dat. On Windows, download and install sftp or WinSCP for file transfer from/to your eniac account.] (4 pts) Use the help() and example() functions to learn how to use the which() and grep() functions. Show how to use these two functions, respectively, to display gene expression values for a gene called "ESR1" in the "ge" data set. [Hint: use the rownames() function to obtain gene names of the "ge" data set.]

April 15: Transcriptome (2): Bi- and Multi-variate statistics

• Learning goal: (1) multivariate clustering analysis; (2) Classification of breast-cancer subtypes
• Lecture slides:
  File:Session-8-small.pdf
• Tutorial 1. Gene filtering

1. Open a new R Studio session and set working directory to R-work
2. Load the default workspace and plot a histogram of the gene expression using hist(ge, br=100). Answer the following questions:
   1. Make sure ge is a matrix class. If not, review the last session on how to make one
   2. What is the range of gene expression values? Minimum? Maximum?

      range(ge); min(ge); max(ge)

   3. Are these values normally distributed? Test using qqnorm and qqline.
   4. If not, which data range is over-represented? Why?

      Low gene expression values are over-represented, because most genes are NOT expressed in a particular cell type/tissue.

3. Discussion Questions:
   1. What does each number represent?

      log2(cDNA)

   2. Why is there an over-representation of genes with low expression values?

      Because most genes are not expressed in these cancer cells

   3. How to filter out these genes?

      Remove genes that are expressed in low amount in these cells, by using the function pOverA (see below)

   4. How to test if the filtering is successful or not?

      Run qqnorm() and qqline()

4. Remove genes that do not vary significantly among breast-cancer cells
   1. Download the genefilter library from BioConductor using the following code

      source("http://bioconductor.org/biocLite.R"); biocLite("genefilter")

   2. Load the genefilter library with library(genefilter)
   3. Create a gene-filter function: f1=pOverA(p=0.5, A=log2(100)). What is the pOverA function?

      Remove genes expressed with lower than log2(100) amount in half of the cells

   4. Obtain indices of genes significantly vary among the cells: index.retained=genefilter(ge, f1)
   5. Get filtered expression matrix: ge.filtered=ge[index.retained, ]. How many genes are left?

      dim(ge.filtered)

5. Test the normality of the filtered data set
   1. Plot with hist(); qqnorm(); and qqline(). Are the filtered data normally distributed?
   2. Plot with boxplot(ge.filtered). Are expression levels distributed similarly among the cells?

• Tutorial 2. Select genes vary the most in gene expression

1. Explore the function mad. What are the input and output?

   Input: a numerical array. Output: median deviation

2. Calculate mad for each gene: mad.ge=apply(ge.filtered, MARGIN=1, FUN=mad)
3. Obtain indices of the top 100 most variable genes: mad.top=order(mad.ge, decreasing=T)[1:100]
4. Obtain a matrix of most variable genes: ge.var=ge.filter[mad.top,]

• Part 3. Classify cancer subtypes based on gene expression levels

1. Discussion Questions:
   1. How would you measure the difference between a one-dimensional variable?

      d=|x1-x2|

   2. How would you measure the difference between a two-dimensional variable?

      d=|x1-x2|+|y1-y2|

   3. How would you measure the Euclidean difference between a three-dimensional variable?

      d=|x1-x2|+|y1-y2|+|z1-z2|

   4. How would you measure the Euclidean difference between a multi-dimensional variable?

      d=SQRT(sum([xi-xj]^2))

   5. To measure similarities between cells based on their gene expression values, is it better to use Euclidean or correlation distances?

      Correlational distance measures similarity in trends regardless of magnitude.

2. Calculate correlation matrix between cells: cor.cell=cor(ge.var)
3. Calculate correlation matrix between genes: cor.gene=cor(t(ge.var)) What does the t() function do?

   transposition (turn rows into columns and columns into rows)

4. Obtain correlational distances between cells: dg.cell=as.dendrogram(hclust(as.dist(1-cor.cell)))
5. Obtain correlational distances between genes: dg.gene=as.dendrogram(hclust(as.dist(1-cor.gene)))
6. Plot a heat map: heatmap(ge.var, Colv=dg.cell, Rowv=dg.gene)

Assignment #8 (Finalized 4/15)

(5 pts) Exercise 4.3 (pg 217) Test significance of gene expressions between the Exp and Ref groups for Gene 2 and Gene 3. Show all R commands, boxplots, and t-test outputs. Are gene expression values significantly different between the Exp and Ref groups? Explain. (5 pts) Exercise 4.4 (pg 224-225). Calculate pairwise correlation coefficients between (1) genes and (2) time points, using the cor() function in R. Calculate and show the heatmap. (5 pts, Read and Respond). Read the "RNA-SEQ" section (pg 234-236). Compare the RNA-SEQ and the Microarray technologies by explaining (1) what is RPKM? (referring to Fig 4.16, the last panel) (2) what is "alternative splicing" and why it could be detected by RNA-SEQ but not by microarray (pg 236, 2nd paragraph)? (3) what is "allele-specific transcription" (pg 236, 2nd paragraph) and why it could be studied with RNA-SEQ but not microarray?

April 22: Transcriptome (3). RNA-SEQ Analysis

• Learning goal: RNA-SEQ pipeline
• Lecture slides:
  File:Session-9－small.pdf
• Read data source GSE32038
• Tutorial 1. Visualize expression levels

gene.fpkm=read.table("../../bio425/data/genes.fpkm_tracking", head=T, sep="\t")
dim(gene.fpkm)
head(gene.fpkm)
fpkm=gene.fpkm[which(gene.fpkm$C1_status=='OK' & gene.fpkm$C2_status=='OK'),] # filter low-quality values
hist(log10(fpkm$C1_FPKM), br=100)
plot(density(log10(fpkm$C1_FPKM)))
plot(log10(fpkm$C1_FPKM), log10(fpkm$C2_FPKM))
fpkm.m=log2(fpkm$C1_FPKM) + log2(fpkm$C2_FPKM)
fpkm.a=log2(fpkm$C1_FPKM) - log2(fpkm$C2_FPKM)
plot(fpkm.m, fpkm.a)

• Tutorial 2. Test differential gene expressions

gene=read.table("~/bio-425-spring-2014/cuff_out/gene_exp.diff", header=T, sep="\t")
dim(gene)
head(gene)
gene.ok=gene[which(gene$status=='OK'),] # filter
gene.sig=gene.ok[which(gene.ok$significant=="yes"),] # select significant genes
plot(gene.ok$log2.fold_change., -log10(gene.ok$q_value)) # volcano plot
plot(gene.ok$log2.fold_change., -log10(gene.ok$q_value), col=gene.ok$significant)

Assignment #9 (Finalized 4/25, Friday 5pm)

(10 pts, counting towards your in-class performance). Transfer the two RNA-SEQ output files ("genes.fpkm_tracking" and "gene_exp.diff", both in the folder "bio425/data/") to your local computer (using "scp"). Make a histogram of the "C2" sample, an MA plot, and a volcano plot. Show all your R commands. (5 pts, Read and Respond). Based on the section "Classification of SNPs" (pg 133-136), define the following terms: (1) SNP, (2) Coding SNPs, (3) Replacement SNPs, (4) Transitions, and (5) haplotype.

April 29: Population Analysis (1)

• Learning goals: (1) SNP analysis (single-locus), (2) haplotype analysis (multiple loci)
• Lecture slides:
  File:Session-10.pdf

Tutorial. Linkage disequilibrium. Install and use the R package "genetics". Understand genotype, individual, SNP site, HWE, LD table

source('http://www.bioconductor.org/biocLite.R')
biocLite("genetics")
library(genetics)
data=read.csv("example_data.csv") # copy the file from "bio425/data" to R-work first
data=makeGenotypes(data)
HWE.test(data$c104t) # Hardy-Weinberg equilibrium
ld=LD(data) # pairwise D', r, and p-values
LDtable(ld) # visualize LD table
summary(lm( DELTA.BMI ~ homozygote(c104t,'C') + allele.count(a1691g, 'G') + c2249t, data=data)) # linear regression
#Challenge: Read the output file from the PERL script above & calculate LD table. Verify with the alignment.

Tutorial. Contingency Table tests (Analysis of Frequencies & Proportions)

fisher.test(matrix(c(9,1,2,4), nrow=2))
chisq.test(matrix(c(9,1,2,4), nrow=2), simulate.p.value=T) # ch-squared test with simulated p value

Assignment #10 (Finalized on Friday 5/2/14 12 noon)

(10 pts). Exercise 3.1 (in Textbook). Quantifying heterozygosity and LD. Using the ten sequences to answer the following questions: Label and count the number of segregating sites (i.e., SNP sites; using a photocopy of the ten sequences) Are there SNP sites showing more than two nucleotide states? [The answer is no, explain why]. Pick any SNP site, can you tell which nucleotide state is a mutation and which is a wild type? [the answer is again no. Explain why] Label SNP sites that are transitions (using "ti") and tranversions (using "tv") List the 10 haplotypes at these sites Calculate allele frequencies at the 2nd SNP site and then obtain the heterozygosity value at this site List counts of four possible haplotypes for the 2nd and 3rd SNP sites (zero for absence of a haplotype). Test significance of linkage disequibrium between these two sites using the Fisher's test in R. Do these two SNPs show significant link disequilibrium? If yes, which haplotypes are over-represented and which are under-represented? Repeat the above for the 2nd and 5th SNP sites. Do these two SNPs show significant link disequilibrium? If yes, which haplotypes are over-represented and which are under-represented?

May 6: Population Analysis (2)

• Learning goals: Case-Control studies for gene mapping
• Lecture slides:
  File:Session-11-small.pdf
• Tutorial: Descriptive genetic statistics of a population. Copy and paste the following BioPerl code as "popstat.pl". Run the code with the "bio425/data/snp.dat" and "exercise-3-1.aln" as the input files. Annotate the statements. Modify the code to print the haplotypes consisting of 1st and 2nd SNP sites. Analysis output (LDtable) using the R package "genetics".

#!/usr/bin/perl           
# Input: CLUSTALW alignment       
# Output: Haplotypes        
use warnings;
use strict;
use lib '/data/biocs/b/bio425/bioperl-live';
use Bio::AlignIO;
use Bio::PopGen::Utilities;
use Bio::PopGen::Population;
die "Usage: $0 <alignment_file>\n" unless @ARGV > 0;
my $file = shift @ARGV; # store filename in variable $file
my $in = Bio::AlignIO->new(-file=>$file, -format=>'clustalw'); # read alignment from the file  
my $aln = $in->next_aln(); # get an alignment object  
my $pop = Bio::PopGen::Utilities->aln_to_population(
    -alignment           => $aln,
    -include_monomorphic => 0,
    -site_model          => 'all'
    ); # create a population object from an alignment object

for my $ind ($pop->get_Individuals()) { # for each individual  
  print $ind->unique_id(), "\t"; # print id
  for my $name ( sort {$a <=> $b} $pop->get_marker_names() ) { # for each SNP site, sorted numerically by positions
    my @genotypes = $ind->get_Genotypes(-marker => $name); #  get genotype as an array
    my $geno = shift @genotypes;  # get genotype at this SNP site
    my @alleles = $geno->get_Alleles(); # get alleles at this SNP site
    my $allele = shift @alleles; # get a single allele
    print $allele;
  }
  print "\n";
}
exit;

May 13: Final Review

• Review slides:
  File:Final-Review-small.pdf
• BioPerl exercises: Modify the above code to (1) calculate heterozygosity at all SNP sites; (2) output 2-loci haplotype frequencies
• Free registration for One-Day (May 29, Thursday) Bioinformatics Symposium at Hunter: Extra credits for full-day attendance
• Teacher's evaluation

May 20: Final Exam

• Due by 5pm, Wed May 21, 2014, @ HN839 (my office)
• Reminder: Teacher's evaluation

General Information

Course Description

• Background: Biomedical research is becoming a high-throughput science. As a result, information technology plays an increasingly important role in biomedical discovery. Bioinformatics is a new interdisciplinary field formed between molecular biology and computer science.
• Contents: This course will introduce both bioinformatics theories and practices. Topics include: database searching, sequence alignment, molecular phylogenetics, structure prediction, and microarray analysis. The course is held in a UNIX-based instructional lab specifically configured for bioinformatics applications.
• Problem-based Learning (PBL): For each session, students will work in groups to solve a set of bioinformatics problems. Instructor will serve as the facilitator rather than a lecturer. Evaluation of student performance include both active participation in the classroom work as well as quality of assignments (see #Grading Policy).
• Learning Goals: After competing the course, students should be able to perform most common bioinformatics analysis in a biomedical research setting. Specifically, students will be able to
  • Approach biological questions evolutionarily ("Tree-thinking")
  • Evaluate and interpret computational results statistically ("Statistical-thinking")
  • Formulate informatics questions quantitatively and precisely ("Abstraction")
  • Design efficient procedures to solve problems ("Algorithm-thinking")
  • Manipulate high-volume textual data using UNIX tools, Perl/BioPerl, R, and Relational Database ("Data Visualization")
• Pre-requisites: This 3-credit course is designed for upper-level undergraduates and graduate students. Prior experiences in the UNIX Operating System and at least one programming language are required. Hunter pre-requisites are CSCI132 (Practical Unix and Perl Programming) and BIOL300 (Biochemistry) or BIOL302 (Molecular Genetics), or permission by the instructor. Warning: This is a programming-based bioinformatics course. Working knowledge of UNIX and Perl is required for successful completion of the course.
• Textbook: Gibson & Muse (2009). A Primer of Genome Science (Third Edition). Sinauer Associates, Inc.
• Academic Honesty: Hunter College regards acts of academic dishonesty (e.g., plagiarism, cheating on examinations, obtaining unfair advantage, and falsification of records and official documents) as serious offenses against the values of intellectual honesty. The College is committed to enforcing the CUNY Policy on Academic Integrity and will pursue cases of academic dishonesty according to the Hunter College Academic Integrity Procedures.

Grading Policy

• Treat assignments as take-home exams. Student performance will be evaluated by weekly assignments and projects. While these are take-home projects and students are allowed to work in groups and answers to some of the questions are provided in the back of the textbook, students are expected to compose the final short answers, computer commands, and code independently. There are virtually an unlimited number of ways to solve a computational problem, as are ways and personal styles to implement an algorithm. Writings and blocks of codes that are virtually exact copies between individual students will be investigated as possible cases of plagiarism (e.g., copies from the Internet, text book, or each other). In such a case, the instructor will hold closed-door exams for involved individuals. Zero credits will be given to ALL involved individuals if the instructor considers there is enough evidence for plagiarism. To avoid being investigated for plagiarism, Do NOT copy from others or let others copy your work.
• Submit assignments in Printed Hard Copies. Email attachments will NOT be accepted. Each assignment will be graded based on timeliness (10%), whether executable or having major errors (50%), algorithm efficiency (10%), and readability in programming styles (30%, see #Assignment Expectations).
• The grading scheme
  • Assignments (100 pts): 10 exercises.
  • Mid-term (50 pts).
  • Final Project (50 pts)
  • Classroom performance (50 pts): Active engagement in classroom exercises and discussions
  • Attendance (50 pts): 1 unexcused absences = 40; 2 absences = 30; More than 2 = 0.

Assignment Expectations

• Use a programming editor (e.g., vi, emacs, sublime) so you could have features like automatic syntax highlighting, indentation, and matching of quotes and parenthesis.
• All PERL code must begin with "use strict; and use warnings;" statements. For each assignment, unless otherwise stated, I would like the full text of the source code. Since you cannot print using the text editor in the lab (even if you are connected from home), you must copy and paste the code into a word processor or a local text editor. If you are using a word processor, change the font to a fixed-width/monospace font. On Windows, this is usually Courier.
• Also, unless otherwise stated, both the input and the output of the program must be submitted as well. This should also be in fixed-width font, and you should label it in such a way so that I know it is the program's input/output. This is so that I know that you've run the program, what data you have used, and what the program produced. If you are working from the lab, one option is to email the code to yourself, change the font, and then print it somewhere else as there is no printer in the lab.
• Recommended Style
• Bad Style

Useful Links

Unix Tutorials

• Oreilly Book for the virtues of command-line tools: Data Science at Command Line by Jeroen Janssens
• A very nice UNIX tutorial (you will only need up to, and including, tutorial 4).
• FOSSWire's Unix/Linux command reference (PDF). Of use to you: "File commands", "SSH", "Searching" and "Shortcuts".

Perl Help

• Professor Stewart Weiss has taught CSCI132, a UNIX and Perl class. His slides go into much greater detail and are an invaluable resource. They can be found on his course page here.
• Perl documentation at perldoc.perl.org. Besides that, running the perldoc command before either a function (with the -f option ie, perldoc -f substr) or a perl module (ie, perldoc Bio::Seq) can get you similar results without having to leave the terminal.

Regular Expression

• A regular expression cheatsheet

Bioperl

• BioPerl's HOWTOs page.
• BioPerl-live developer documentation. (We use bioperl-live in class.)
• Yozen's tutorial on installing bioperl-live on your own Mac OS X machine. (Let me know if there are any issues!).
• A small table showing some methods for BioPerl modules with usage and return values.

SQL

• SQL Primer, written by Yozen.

R Project

• Install location and instructions for Windows
• Install location and instructions for Mac OS X
• Install R-Studio
• For users of Ubuntu/Debian:

sudo apt-get install r-base-core

© Weigang Qiu, Hunter College, Last Update ~~