NGFN Leukemia Bioinformatics Resource

Coordinator:    Dr. Claudio Lottaz
Institution: Institut für Funktionelle Genomik, Universität Regensburg
Our focus is on studying gene expression profiles from patients and model systems. Also, we integrate data within and across sub-projects of the consortium. While gene expression data makes up the backbone of our studies, genomic data (changes in the sequence and copy number of the DNA) provide valuable additional insight into the molecular causes of leukemia. We found the following novel insights corresponding leukemia together with partners from the consortium:

Together with partners from sub-project SP04 from the University of Heidelberg, we found a striking correlation between messenger- and micro-RNA. (Very short and rarely studied RNA.) For many genes, we identified groups of miRNAs that can predict its activity. Vice-versa, we found groups of mRNAs that reliably predict the activity of a miRNA. The relevance of these connections in leukemia need further analysis and may provide novel hints towards the function of miRNAs that are particularly active in leukemia.

Together with partners from sub-project SP14 from the Charité in Berlin, we analysed the DNA copy number changes in relapses of leukemia patients. In cancer cells from patients, DNA copy numbers have been measured on the whole genome at diagnosis, after remission and at relapse. We found that many patients have similar DNA changes which often reappear at relapse. Such repetitions indicate that relapses are often caused by the same cancer cell.

Together with partners from sub-project SP01 from the University of Ulm, we detected pathological changes at single genome positions in cancer cells from leukemia patients that may influence their course of disease. We have found a number of DNA differences in the analysed patients that improve the characterization of individual tumors and clarify underlying general mechanisms. For instance, mutations that influence chromosome organization are particularly frequent in leukemia.

In order to collect data from experimental and clinical projects in a common analysis, we have developed an integrating analysis method. For ethical and technical reasons, sensitivity of an experimental drug cannot be tested in patients. Therefore, cultivated cancer cells from patients are used here to predict the drug’s effect in patients. For this aim, our novel gene expression analysis method called Guided Clustering compares correlation structures instead of direct gene expression profiles, such that we can analyse datasets from very different cellular contexts. So, we can deduce correlation structures from cancer cells from patients which estimate the sensitivity of cancer cells to a drug. Thus, our extended Guided Clustering can predict the sensitivity of patients and has therefore the potential to improve personalized medication.

Figure legend:
Connections between compounds of the Connectivity Map and leukemia datasets. A green edge indicates direct similarity, red reversed similarity, blue similarity in both directions, black indicates that results from replicate experiments were contradictory. Some compounds connect to many datasets indicating similar changes in gene expression between compound treatment and the leukemia datasets.

Further Coordinators:
INTRANET (Members login)