I need to find patterns experimental data.

The columns are "experiments" which are chemical treatments for growth experiments. The rows are individual gene names, the values are a fitness-defect score, which reflect the genes contribution to growth.

I would like to find patterns that are reflected across all experiments using some type of PCA or clustering. I have been trying to use sklearn but have not been successful in applying a model.

The data looks like:

gene     SGTC_1                   SGTC_2                 SGTC_3 
YAL002W  3.56420220283773        1.80774301690328       0.431491057210906
YAL004W -0.885645399324204      -1.76020417788351       0.883034190306176


There are 4000 rows for genes and 30 columns for experiments.

Any suggestions would be greatly appreciated.


PCA is a dimensionality reduction algorithm - it projects your high dimensional data onto a lower dimensional plane. This is useful for either visualisation (if you reduce to 2 or 3 dimensions to plot), or for training machine learning models.

You say you want to find patterns in your data - I’m not quite sure what you mean by this. Do you want to visualise your data, train a model on it and make some prediction, or something else?

To visualise high dimensional data, you could use either the tSNE (t-stochastic neighbour embedding) algorithm, or PCA.

Depending on the type of data you have, you can “find patterns” in different ways.

If your data is unlabelled (you don’t know the classes of each sample or there is no dependent variable) you can use unsupervised learning algorithms such as K-means clustering, K nearest neighbours, Gaussian mixture model. If your data has dependent variables, depending on whether your dependent variable is categorical or continuous, you could use classification algorithms for the former and regression algorithms for the latter. Classification algorithms include logistic regression or decision trees. Regression models include linear regression.

  • $\begingroup$ If you feel I answered your question, don’t be afraid to upvote my answer and mark it as correct :) otherwise, please let me know if I can explain any more and I’ll do my best $\endgroup$
    – PyRsquared
    Mar 6 '18 at 20:08
  • $\begingroup$ Your comment was very helpful. I have been asked to visualize the data and eventually make some predictions. My data is unlabeled. I will try K-means clustering. $\endgroup$
    – mplace
    Mar 7 '18 at 21:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.