Replies to E. Manduci’s Comment

March 22, 2009

Thanks Elisabetta-  pls see below for answers to your questions.I think a newer version is coming your way in a few days. In the mean time, you can use this one to play with the actual calc and see how it behaves, nothing has changed there.




—–Original Message—–

From: Elisabetta Manduchi

Sent: Friday, March 13, 2009 2:09 PM

To: Valtchinov, Vladimir I.,Ph.D.

Subject: Re: Fwd: Relevance network software



Hi Vlad,

Chris forwarded me your message below and the software which I’ve installed and am testing on a pilot dataset.I have a few questions to make sure I’m interpreting the output from ‘Start Calculation’ and its parameter settings correctly (I’ve looked at the manuals you pointed us to, but I didn’t see answers to these there):


1. Num of bins N: I imagine these are used to compute entropy and mic. How are the data binned? First of all, do you bin separately the range

of each gene into N bins (this is my guess) or do you bin the range of values over all genes?

[Vlad Valtchinov] 2 histograms are generally needed, 1-d and a 2-d (Contingence Table).

Both are global, in a sense that you first take min and max of all the entities.

Second, do you bin into equal width bins? (As oppoed to equal frequency, which is the default e.g. for the minet R software).

[Vlad Valtchinov] equal width


2. Permutations: given a specified number of permutations B, for each i=1,…, B do you separately permute the data for each gene (this is my guess) or do you permute the whole data matrix?

[Vlad Valtchinov] for each gene.


3. In the output from Start Calculation, what is the entropy metric computing: the entropy of id_ref1, that of id_ref2, the joint entropy?

[Vlad Valtchinov] Joint Entropy or the Mutual Information Content (MIC).

Now, for some Entropy based filters we do compute the entropies of the individual entities (genes) but I am not sure if we are not suppressing the output right now…


4. Does the value under permuted<Metric> represent the *average* value for that pair of genes over the B permutations?

[Vlad Valtchinov] it is the max of all permuted values for MIC metric.We are now adding a printout statement to collect all permuted values for each metric(as a comma-separated string) to facilitate analysis looking at the distribution of these.


5. Does the min<Metric> represent the min over the B permutations and similarly for max<Metric>?

[Vlad Valtchinov] yes.


6. I’m running this on a pilot small dataset consisting of RT-PCR data on 47 genes over 15 samples (time points), so I’m not in the situation in which valid intervals need to be used. I see that the time stamp column gets populated. What are those values and are they used in any way?

[Vlad Valtchinov] just a time stamp when the result was produced.


Moving on to the Create Graph step, a fewquestions:


1. How/where is the rank column used?

[Vlad Valtchinov] we used to want to put the rank of a list of correlated gene pairs on the yEd graph. For the time being, pls abstain from this feature,we are adding some more functionality and making it clearer in the new release,coming up shortly.


2. What does checking ‘Neg. Metric clr.’ do?

[Vlad Valtchinov] oops, ‘Negative Metric Color’; the color picker selects that as well.


3. It appears that when I try to change the color to display negative valued cells for the selected metric, the change doesn’t get applied (i.e.

it stays red), just thought to let you know.

[Vlad Valtchinov] was a bug, it is fixed now.

Welcome to the RelNet Blog!

November 12, 2008

Welcome to New Atlantic Technology Group (NATG), LLC’s blog on Relevance Networks, Correlation Analysis and Parallel Computation.

This blog is dedicated to Relevance Networks computation and implementation issues which is the core methodology behind the i2b2 hive’s Correlation Analysis Cell.

The i2b2 hive is the scalable computational platform the National Center for Biomedical Computing Informatics for Integrating Biology and the Bedside (i2b2) has been set out to develop and disseminate.

The Correlation Cell was developed by i2b2 and NATG, LLC under a contract supported by a grant from Partners HealthCare System, Inc. It is an open-source effort which distribution is governed by the i2b2 Software License Agreement (“Software License”) (see it here).

NATG is committed to providing a basic level of support as part of the open-source distribution philosophy of i2b2. Additional, commercial support and enhancements options are available for interested organizations.

Please create a new post describing the problem you are experiencing or look for other similar posts.

Thank you.

Vlad Valtchinov, PhD