Prerequisites Overview

The biggest prereq, especially for people using this package internally, is obtaining a protein structure.

When a structure is not available in the PDB, you need to do two things: (1) decide which fragments of proteins to model, and (2) model those fragments.

To choose which fragments to model, I wrote some python scripts that I'll go over in future sections.

To model fragments, we will use two tools: I-TASSER (for template-directed modelling) and QUARK (for de-novo modelling). Both of these tools are located on O2 at /n/groups/drad/.

Quick Reference

A quick reference for important folders and notes.

Naming Convention

All headers are named "Protein_FragmentID_LocationStart_LocationEnd", where location is optional.

• O15347_PF00505_HMG_box_93_161 is a PF00505_HMG_box pfam from the O15347 protein located from residues 93 to 161, inclusive.

• O15347_full_protein is the full protein sequence for the O15347 protein.

PDB files are nested in folders with their protein and fragment identifiers.

• The .pdb models for O15347_PF00505_HMG_box_93_161 are located at /n/groups/drad/all_pdb_models/human/O15347/O15347_PF00505_HMG_box_93_161

• The .pdb models for O15347_full_protein are located at /n/groups/drad/all_pdb_models/human/O15347/O15347_full_protein

Modelling utility scripts

Utilities to generate pfam-aware fragments to model: Modelling_Pfam_Selection, this git repository.

Utilies to run I-TASSER in high-throughput: /n/groups/drad/I-TASSER4.3/BatchScripts.

Utilities to run QUARK in high-throughput: /n/groups/drad/QUARKmod/BatchScripts.

Utilities to move, archive, and run modelling files: /n/groups/drad/julian/Modelling_Monitoring.

Utilites to track the progress of runs: /n/groups/drad/all_pdb_models/progress_logs.

Modelling pipeline

1. Decide what sequences to model, usually using the scripts Modelling_Pfam_Selection in the git repository.

2. Create csv file in format "header,sequence"; upload to O2 in a new BatchJobs directory. csv file for I-TASSER should have sequences under 1500aa, csv file for QUARK should be under 800aa.

3. Run csv_to_fasta.sh, make a main and scripts folder within the BatchJobs directory.

4. Run createITasserJobScripts.sh and/or createQuarkJobScripts.sh with -i main -o scripts.

5. Run runJobScripts.sh with -s scripts.

6. Each day, monitor job progress. Use /n/groups/drad/julian/Modelling_Monitoring/transfer_archive_models.sh to centralize all models and archive jobs, and then /n/groups/drad/all_pdb_models/progress_logs/Makefile to generate a summary of the completed jobs.

7. Many jobs will fail. To check and re-run failed jobs, use /n/groups/drad/julian/Modelling_Monitoring/jobrunner_helper.sh with appropriate parameters.