The genomics era, which began with the sequencing of human DNA in 2003, was the cornerstone of a revolution -- for the first time genetic differences between different individuals and between humans and other animals could be measured. A key part in this revolution is the ability to perform molecular characterization of tissues using sequencing; one can compare a patient with a diseased tissue to a healthy one and discover the molecular differences that underline the disease. There is, however, one major limitation to all sequencing technologies - they all begin with dissociation of the tissue and extracting the molecules of interest from it. This means that we completely lose the spatial information - the location of the molecules inside the tissue. This is unfortunate as spatial location is of great importance - classic examples are in brain tissues, where a difference of less than a micron in a molecule's location can cause disease, and in cancerous tissues.

 

To overcome this limitation our lab develops technologies that allow in situ biology - that is, the ability to extract information about the molecular content of tissues inside their original location. The main challenge of in situ biology is how to perform basic reactions, like signal amplification and sequencing, inside a dense tissue environment? We enable in situ biology by a multidisciplinary approach - merging tools from chemistry, biology, microscopy, engineering, image analysis and data analysis. 

​

The work begins with chemistry - we create a gel inside the tissue and make the tissue physically bigger (see video below).

​

We then use molecular biology - the previous step creates space inside the tissues - we use this space to diffuse enzymes into the tissue and increase the RNA/DNA signal by a thousand fold.

​

The next step is microscopy - the tissue is physically situated on a microscope, each one of the bases -- A, C, T, and G -- is labeled with a different color and thus the sequence of the RNA/DNA can be identified. Since the tissue is physically expanded, we achieve super-resolution with a fast spinning disk microscope. 

​

We then use automation - we built an automated system where a computer controls both the microscope and a fluidics apparatus that automatically replaces all the enzymes needed for the sequencing reaction. The microscope is basically converted into a sequencing machine.

​

Then comes the image processing step - the microscope produces huge image files (terabits per sample) and therefore we use highly sophisticated algorithms to process it. The image processing must be extremely accurate - if a certain point in space moves during the experiment, it might mean that we cannot identify a particular gene. The output of image processing is the location of each gene within the tissue along with its sequence.

​

And finally comes the data analysis step - the resulting data is a unique combination of 1D information (sequence of bases) embedded inside the 3D space of a tissue; we are developing machine learning tools to analyze it, starting with fundamental questions such as: where are the genes located within the tissue? does gene location change as a result of a disease? As the typical number of molecules we characterize inside a given tissue is in the order of millions, the data analysis poses significant challenges.