My docking results on the same grid and the same machine are different. For the same grid both the Windows and Linux systems also give different a docking energy. I read that the system generates this random seed number that might be a possible cause of differences. Could someone explain me what these random seed numbers are and how they affect the docking results?
Hi Ashmit,
It's because 'random' numbers in computer lingo, is not random at all, in fact it is more like a list of random numbers. When you the ask for some random numbers, the computer will return a list of numbers from the list of 'random' numbers starting at a given position in the list of random numbers. This given position is known as the seed.
If you use random numbers when generating some results via e.g. a model, you can choose a constant seed, this way the result will be reproducible, otherwise you'll get a slightly different (presumably) result each time you run the analysis.
Hope it helps :-)
To whomever down-voted my answer, I realise that my answer is not entirely correct, this was my intention based on how the original question was formulated. From a learning point of view, It makes no sense to give rocket-science answers to "simple" questions. We all have to start somewhere...
The purpose of randomness is to get different results. I don't have any knowledge about autodock/vina, but I will comment on random number generators (RNGs). Many people seed a RNG with the current time to have a different outcome of the simulation every time they run the simulation. Actually, this is not a good idea because RNGs don't produce true random number sequences with just any seed.
The idea of Monte Carlo simulations is to use a RNG to produce a random number sequence. A number sequence is called random if it passes certain statistical tests. The whole idea of scientific experiments is to have reproducibility. Using different unknown (at least to you, what you call random) seeds for the simulation is not scientific. Some simulations need "random" numbers. That is why we use RNGs in Monte Carlo simulations. But, you should use a known seed to produce a random number sequence and thus make the results reproducible. Some more information about this can be found in the documentation of the Tina RNG library.
Welcome to one of the unfortunate secrets of computational chemistry.
The problem might be nothing to do with "pseudo-randomness" but may lie in using two different platforms for the calculation. Windows and Linux use different compilers etc. and therefore will suffer from different rounding errors during the floating point calculations that Autodock performs Even if you gave the same random seed to Autodock on Windows and Linux, and therefore have the "same" starting conditions for the docking calculation, you are very likely to get different answers from the two platforms due to differences in handling minimisation calculations and rounding in floating point numbers.
There are two very interesting papers by Chris Williams and Miklos Feher on the influence of tiny differences in input and differences in platform on docking calculations. You should read these papers
- Badges
- Science topic