conda is not found on travis - WTF?

See eg https://travis-ci.org/Becksteinlab/MDAnalysis-workshop/builds/140066856

The talk will contain two sections and will be followed by the hands-on tutorial sessions #2

Introductory talk

MDAnalysis library

@orbeckst (10 min)

1. aims
2. open source (contributors!)
3. big-picture overview
4. basic example as introduction
   • loading from Universe(topology, trajectory)
   • iterating through trajectory
   • accessing coordinates; automatic unit conversion
5. example of interaction with the scientific python eco system: LeafletFinder

MDSynthesis

@dotsdl (5 min)

1. the problem: heterogeneous simulation data storage
2. the solution: hdf5 / datreant
3. example

The basic overview will be covered in the talk #5 . We will use interactive Jupyter notebooks to develop the problems. We will use clean notebooks as guides; participants can use the clean notebooks afterwards.

The following broad concepts should be covered in about 1:40h:

Tutorial

Session 1 (9:45 - 10:30)

@dotsdl

• fundamental objects
  • Universe - loading
  • AtomGroup and Atom
    • properties and methods
    • Advanced: topology objects
    • Example: Ramachandran plotter (can use ag.phi(), ag.psi() or possibly topology objects)
• selections
  • basic select_atoms() language (note: PBC)
  • Example: collective variables (AdK domain angles)
• introduction to trajectory reading
  • iteration pattern
  • slicing
  • Timestep

Session 2 (11:00 - noon)

@orbeckst

• trajectory processing (reading)
  • random access
  • dynamically updated trajectory data (AtomGroup.positions)
  • Examples
    1. get CV timeseries and plot (AdK angles from DIMS)
    2. Ramachandran analysis of trajectory (2D histogram)
    3. velocity autocorrelation function of e.g. ions or water with scipy.fft and diffusion coefficient
    4. segment trajectory based on order parameter (e.g. detect conformational transition in equilibrium AdK trajectory)
• modifying coordinates
  • translation and rotation
  • RMSD superposition
  • Example: AdK domains
• trajectory processing (writing)
  • writing coordinates (Writer)
  • Example: format changes (Multi PDB/XTC) and concatenation (ChainReader)

Bonus

• interfacing with other packages
  • Examples:
    • implement LeafletFinder with networkx and analyze bilayer formation in a CGSA-MD simulation (@orbeckst)
    • pandas with MDAnalysis (@dotsdl )

As a last resort, people can also use VMs (in vm subdir) but installing VirtualBox and Vagrant comes with its own set of problems. Nevertheless, the VMs are useful for us for testing installation recipes (at least different Linux distros).

I have a suspicion that the VMs that I built do not work anymore because the URLs of the boxes are not valid anymore. It might be sufficient to replace them with "bento"-based boxes from https://atlas.hashicorp.com/bento/ or search for official boxes at https://atlas.hashicorp.com/boxes/search .

The ones that should work/should be tested:

• Debian
• Ubuntu
• CentOS

We need to give users instructions for how to install MDAnalysis on their own machines before the workshop starts. We rather have the participants install their own instead of providing cloud instances because experiences shows that users need to know that they can set up the software themselves.

We support Linux and Mac OS X.

We need

• web page with installation instructions (can be a wiki page here)
• repository or server with data (e.g. @orbeckst's unlimited ASU Dropbox?) --- see #3

What kind of installers should we support?

• pip (+ apt-get) — see the VM recipes
• conda (see the MDA travis.xml)

EDIT

• removed "script for downloading bits and pieces" — can be added to #3 if necessary

• get trajectories from @sseyler
  • AdK equilibrium in solvent (~200 ns) (TPR+GRO+XTC or PSD+DCD)
  • AdK DIMS 3 closed-to-open, 3 open-to-closed (1 PSF, 6 DCD)

• download site
  • public link dropbox: CECAM_Workshop/MDAnalysis/
  • backup: c3potato (TODO)

• set up web page