NCBI GEO database

Transcriptome Browser: http://tagc.univ-mrs.fr/tbrowser/

ArrayExpress is excellent: http://www.ebi.ac.uk/arrayexpress/

Both GEO and ArrayExpress have R Bioconductor support, making it easy to analyse the data.

There are no repository data bases that offer true, in depth analysis options.  Geo2R at the NCBI web site has some significant limitations and caveats when it comes to analysis and is primarily intended for exploratory analysis to see if a particular series of arrays is comparable before you go any further.  For one thing, NCBI's Geo2R will timeout after 10 minutes whether the analysis is complete or not.  Also Geo2R assumes you have checked the data to ensure the comparison you are setting up is actually valid - it does not check if that is true or not, it just runs the query you set.

Ultimatly, to analyze any of the publicly available microarray data or NGS sequence data in depth, you really need to download the raw data and analyze it yourself.  This can be done with tools in BioConductor using R, or any number of other open source and commercial packages.

For E. coli specifically, you should look at http://expression.porteco.org.  ExpressDB is now defunct I gather, but the datasets can be downloaded still - http://arep.med.harvard.edu/ExpressDB/ and there is ArrayExpress at EMBL - http://www.ebi.ac.uk/arrayexpress/

For most species you will find you cannot count on conveniently finding everything in one or often even a few web sites.  You have to dig around through several sites and databases to fully mine the available data.

Hi Michael B Black

Geo2R is great but it is only for normalized data like Affymetrix, and it can not normalize your data

Here is a very similar question to what you asked, which I thought could be helpful: https://www.biostars.org/p/52866/

I think Sequence Read Archive (SRA) could also be an option, but supposedly it is not as comprehensive as GEO itself.

Good luck!

Thank you Jin Gu, Touati Benoukraf, Ignat Drozdov, Michael B Black, seyed morteza Razavi, Bilgenur Baloglu,....  for your kind opinions. I tried the suggested databases and it seems like l have to download the data from different databases to gather the maximum possible information. 

Currently, I am using Bioconductor R for microarray data analysis. Please let know if you have any idea about other open source or freely available softwares for easy and extensive analysis of Microarray data. It will be even better if it is capable of handling data from different platforms.

Warm Regards

Ashish

Ashish, it is very difficult to simultaneously analyze expression data from different platforms.  Different platform raw data is simply not something you can combine, normalize and analyze together.  Each platform has its own optimal normalization strategy and those will be different and incompatible between platforms.  You also have to deal with the differences in probe annotation and annotation matching between platforms.  Note that ideally, you should be matching probe by probe based on the specific genomic transcript interrogated - Affymetrix has such best match lists for cross-array comparisons of many of their own different chip sets, but going from Affy to Agilent or Illumina or whatever is much more involved as the probe level matching likley has not been done or at least is not publicly available anywhere.

If you try to combine annotation matched subsets of different platforms using normalized data, and then apply a second round of normalization to the merged dataset, the results are often disappointing.  The problem being you end up reducing the variance of all the data so much that you will have nothing left to differentiate anything at all.

R & BioConductor have tools for every type of platform (LIMMA, for example, is suitable for analysis of normalized data from any platform).  However, if you want to try and perform a meta-analysis of data from different platforms, you will have to research some much more involved statistical normalization schemes to deal with the inherent problems of trying to combine such data.  Such analyses are difficult and complex.

There are strategies for meta analysis of data from a single platform (e.g. MAAMD for Affymetrix data).  Have a look also at http://www.hindawi.com/journals/tswj/2012/989637/

More often what people do is analyze experiments independently, and then use comaprative statistics to look for commonality or differences across experiments and/or platforms.  For example, if you have estimated fold change from a human Affymetrix experiment and estimated fold change for a comparable human Agilent experiment, you could use a non-parametric tool like RankProduct to compare the two gene lists (z-scores can aslo be useful for comparing across experiments).