Mitocheck is the name of a molecular biology online database that was created by the same consortium. Funding was provided under the European Sixth Framework Programme for Research and Technological Development (FP6, 2002-2006 ). Mitocheck provides experimental data in the form of videos and graphics available that were recovered from the culture changes in HeLa cells. One goal is to retain the phenotypic characteristics of cell division (mitosis ) of Loss -of-function mutations in this cell line for each gene in the human genome, into categories and so the possible function of each gene product (protein) in cell division to determine. The usually carried out for individual genes work steps have been automated in this project and parallelization, which is a novelty in such projects. By Mitocheck unknown genes of cell division were initially 600 partially pre- assigned and are identified in a follow-up project 100 protein complexes.

Standard experiment

To determine the function of a gene product in HeLa cells during cell division, first the respective gene " sedated ". This is done by adding appropriate siRNA for cell culture. It is stored in the ideal case, each mRNA molecule of which is transcribed from the gene and prevents further translation. Thus there is no or very little corresponding protein. This protein is normally an important function during cell division, one can see this in the consideration of dividing cells in a microscope, especially when histones of the cell were stained with GFP. Lack of enzymes or structural proteins can lead to altered flow or complete defect in cell division.

Automation and parallelization

As part of the project Mitocheck several automation steps were first designed and initially tested in a pilot project with 49 genes.

RNA interference microarrays

The microarray technology allows to perform hundreds of experiments simultaneously. The Mitocheck project used self-generated RNA interference microarrays, where previously automatically synthesized siRNA were fixed positions on the individual ( delimited from one another ) and then covered with the cell culture. Hereinafter the siRNA diffuses into the cells and binds to the appropriate mRNA.

Time-lapse microscopy

To follow all the steps of cell division, it is apparently necessary to observe the cells at least one cell division period is long. Not only to automate this process, but also to parallelize Mitocheck developed a fluorescence microscope, which was able to map the positions of several hundred to a microarray in each case at an interval of several minutes. The individual images of each position result in a time-lapse recording of the event. This is sufficient to determine changes from the norm and into categories.

Image processing by machine learning

To automate the image processing, put the Mitocheck project a methods that are commonly referred to as machine learning. Here were three problems to solve: the localization of individual chromosomes by means of a locally adaptive thresholding, morphological classification of chromosomes using previously manually trained support vector machines, and unsupervised phenotypic classification of the entire time-lapse recording by cluster analysis.


The decommissioning of about 21,000 genes produced approximately 190,000 time-lapse videos showing the 19 million cell divisions. Below 1.9 billion cell nuclei were classified. A total of 1,249 genes resulted in a deviation in cell division. To eliminate false positives, a second run was carried out with 1,128 of these genes and modified siRNA, which eventually resulted in 572 genes that showed consistent results.