I have done DGGE gel analysis for three different genes from ten rhizosphere plants. I have got the data on band intensity and SW index and I would like to do Principal Component analysis (PCA) with SW index alone for ten samples along with all three genes in a single PCA analysis. Is it possible to do PCA with the available data and also any software that can be easily used for the same. Need help in this.
Hi Baskaran,
If you're trying to do PCA on a series of images, check out: http://biochem.missouri.edu/trend/
If your data is a numeric matrix you can use my webserver: https://beamerlab.shinyapps.io/software/
The webserver expects data in the format 'ID | Variable1 | Variable 2| etc' with the first row of the CSV being labels, first column being identifiers, optional 2nd column of groupings, and the following columns all being numeric data. If you don't have groupings in your 2nd column, make sure to change the slider to "2".
Good luck!
Hi Baskaran,
To get started, you need to check your data on the conformity requirements of principal component analysis. It relies on four major assumptions:
1. Linearity. This means that the only interaction among different signal sources is that they add together. If the strength of a combined signal were the product of the strengths of contributing signals, for instance, this would be a non-linear interaction and PCA would not work.
2. The interesting dynamics have the largest variances.
3. Mean and variance are sufficient statistics. Since PCA is designed around the covariance matrix of mean-centered data, the only statistics it considers are the mean and variance. If the data cannot adequately be described by its mean and variance (e.g. it is not Gaussian or exponentially distributed), PCA will be inadequate.
4. Orthogonal components. This is a particularly strong assumption made by PCA. It is this assumption that allows PCA to be computationally straightforward, but is not a realistic assumption in many data sets.
For training, you can try a free program the PAST - a fairly simple program to use, which implements coordination methods, including the principal component analysis. http://folk.uio.no/ohammer/past/
Please see also the discussion below
https://www.researchgate.net/post/What_is_the_best_way_to_scale_parameters_before_running_a_Principal_Component_Analysis_PCA
Best wishes, Denis
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS