01. Modelling

#!/bin/sh

## Pathway Commons.

# This is where it all starts.

# Start with a blind search of Ubiquitin and NF-kB Alternate.

# Browse through pathways and download the interesting ones.

firefox http://www.pathwaycommons.org/pc/record2.do?id=517194

wget http://www.pathwaycommons.org/pc/downloadBioPax.do?id=517194

wget http://www.pathwaycommons.org/pc/downloadBioPax.do?id=486346

wget http://www.pathwaycommons.org/pc/downloadBioPax.do?id=486368

# Finally

# Download the Alternate Nf-kB, Ubiquitin, TRAF pathways

## STRING DB

## Cytoscape

# This study is resource intensive. Increase the memory footprint of JVM in either Cytoscape.vmoptions or on cmdline.

java -DXmx 5G -DXss 1G cytoscape

# Start with a blind search over AgilentLiterature Search addon.

# Put in similar queries as provided for Pathway Commons.

# Take the new results and go back to Pathway Commons to find respective pathways.

# Download the biopax.xml files.

# Load the networks in Cytoscape.

# Finally

## NCBI

## MUSCLE Alignment

# Get an initial alignment to prune false positives, if exists.

muscle -in ./seq.fasta -out seq.afa -tree1 tree.phy -anchors -log sq.log

# Get an html output file for easy viewing. Only suitable for small seq.

muscle -in ./seq.fasta -out seq.afa.html -tree1 tree.phy -anchors -html -log sq.log

# Refine the alignment

muscle -in seq.afa -out refined.afa -refine > sq.m.ref.out

# View the dendogram

perl treedraw.pl > fig7.svg

# Finally

# If some sequences seem to be too much out of way in the alignment and dendogram, then prune them and move on to next step.

## CLUSTAL MSA

# Create the alignment file in stockholm format to be used by HMMER

clustalo -i seq.fasta --distmat-out="seqDist.50" --guidetree-out="seqTree.dnd" --full --cluster-size=3 --clustering-out="seqCluster.cls" --trans=3 --posterior-out="seq.pos" --percent-id --outfile=sq.st --outfmt=st --resno --output-order=tree-order --iter=1000 --threads=4 --log="seq.log" -v --force

# Visualize the dendogram for differences from muscle o/p, if any.

perl treedraw.pl > fig7.svg

# Finally

# Match the results with MUSCLE o/p. If the results are are too different then see if some parameters need to be changed.

# Take sq.st alignment file and move on to next step.

## HMMER Search

# Build the hmm profile

hmmbuild sq.hmm sq.st

# Search the swissprot database with the profile

hmmsearch ./sq.hmm /run/media/aamijaninaa/fedora/home/bio/fromHome/NewFolder/database/swissprot > sq.out

# Perform PSI-BLAST iterative search

jackhmmer ./sq.st /run/media/aamijaninaa/fedora/home/bio/fromHome/NewFolder/database/swissprot > sq.psi.pfam.out

# Finally

# Study the sq.out and search for matches having E-Value between e-60 and 0, Bias = 0 and Score > 300

# Study the sq.psi.pfam.out and search for matches with similar parameters.

# Further the search with while studying the MSA and picking up sequences having similarity around the motifs of interest.

## RCSB Search

# Note the biological units of interest from the previous steps.

# In case of a duplicate structures found in the PDB db, look into Remark 100 for date of entry. Select latest entry

# Study Remark 350 for structure details.

# Study Remark 465, 470, 500 for missing residues and atoms and peculiar torsion angles.

# Specify the required sequences in the perl script.

perl downloadPDB.pl

# Finally

# Verify that all the sequences requested were downloaded.

# Sometimes the file size varies. Make sure that the complete PDB file was downloaded in such cases.

## MODELLER

# After deciding which sequences to consider for modelling and downloading their PDB structures, move on to Modeller

# Iteration 1: Take only the target sequences and perform all the steps in @Structure Alignment

# Iteration 2: Take the structures reported in build_profile.log and append them to the end of existing seq.fasta and perform @Structure Alignment one more time.

# Iteration 3: Decide final templates and model.

# Structure Alignment

# Running the program will

# 1a. Convert the i/p fasta sequence into PIR format.

# 1b. Compress the PDB95 database. Further runs may skip step 1b.

# 2a. Read the binary sequence database.

# 2b. Read the PIR format alignment file.

# 3a. Create a profile alignment.

# 3b. Scan the SWISSPROT sequence db to pick homologous sequences.

# 3c. Create sq.pir MSA.

python3 build_prof.py

# Finally

# seq.pir file obtained is a structure alignment of the i/p sequences. That is why it complains there are some distant seq.

# Match that str aln with the seq aln files obtained from MUSCLE and CLUSTAL.

# Study the differences. If they can be explained, then fine else repeat everything.

# Study the dendogram and select template.

## Intermission

# At this point, you have a structurally homologous subset of templates of interest.

# Repeat everything to see if there is anything else that could be considered.

# Model Building

# Read in the template structures and add to alignment.

# Create a structure alignement of templates with target model.

# Refine the loops and create the final model.

# Evaluate structures using DOPE Score.

python3 build_model.py

# Finally

# Study the DOPE Graph and DOPE Scores.

# Select a final model

## Manual Homology Modelling

# Take the best hit based on GA and DOPE Scores and model it manually in Swiss PDB and Pymol.

pymol best_hit.pdb