Our tRNAome bioinformatics identifies a unique tRNA pair that repeats at high frequency in the target protein’s sequence, but is not frequently used in the background human proteome. Amazingly, 83% of known proteins have a tRNA pair with a repeat factor of 30 or more against the average background. When the selected “signature” pair is used in the screening system, a light pulse (FRET) is generated whenever the pair sits in a ribosome. This will happen at high frequency for the target, but not for other proteins, enabling us to visualize the target protein’s translation. With this capability, we can see when, where and how much of the target protein is being made within the cell in real time.
To monitor overall protein translation, all 46 types of labeled human tRNAs are used in the screen. This generates a light signal on each and every pair, creating a fascinating light map of the overall protein synthesis in the cells. The combined use of total tRNA and a specific pair of tRNA enables differentiating global translation inhibitors and those that are specific. Only compounds that show activity in inhibiting translation with the target’s specific tRNA-pair, but do not show activity on global translation (measured with total-tRNA), will be selected for further analysis.
Translation image analysis
Our proprietary image analysis extracts a large set of features from the millions of images of protein translation. Each feature represents a meaningful biological information item that is used in the generation of a translation signature profile.
We use a data driven approach with proprietary data analysis methods that have been custom built for analyzing protein synthesis images:
- The most informative features that reflect specific biological activity are selected.
- Features are independently selected computationally and cross-validated biologically.
- Compounds with known mechanism of action are tested, based on a test plate with known compounds.
- The biological and pharmacological relevance of the data is ensured by our data analysts working hand in hand with the biologists.
- "translation signature profiles" are created as a phenotype.
- cFRET features show high scores & reproducibility across different screens.
- Features like number of spots, brightness, localization and size are used as good indicators of a compound’s effect on the protein synthesis process.
Big data algorithms
We developed proprietary algorithms to analyze the huge amounts of data in millions of protein synthesis images generated in a typical screen. These include:
- Machine learning
- Dynamic thresholding
- Designated algorithmic approach / project : for each screen, some new elements in the analysis are specifics for cell-lines and indication
In the image: Each cell population defines a volume in space. This multi-dimensional approach allows us to point out possible hits regardless of their distribution.
Hit selection, validation and de-risking
Analysis is done in several iterations by applying multiple secondary assays built around our technology. We analyze the impact of hit compounds on overall protein synthesis, eliminating compounds that are toxic and that interfere with the synthesis of many of the cell's proteins or that shut down the translation machinery itself. Additional translation selectivity assays were developed, which monitor the effect of compounds on groups of proteins of interest or specific proteins along a pathway.
PSM (protein synthesis monitoring) ASSAYS
PSM analysis of compounds in target cells (at least two different cells)
PSM analysis using cell free translation systems
PSM signature of safe drugs
Effect on mRNA levels and localization
Monitoring general protein synthesis: newly synthesized proteins
Monitoring translation regulatory proteins
Ribosome footprint coupled to proteomics
Compound safety: Translation impact analysis- Library of approved drugs is screened against PSM translation panel
Comparing the compound’s translation impact profile against the profiles of known, safe drugs
Di-tRNA combinations representing signature pairs of critical pathways
Several cell types: epithelial (kidney, liver),fibroblasts and neuronal cells
Two different conditions: with or without starvation (starvation is known to have an effect on translation)
Anima's cloud-based system is hosted on Amazon (AWS) in a private secured zone, using the highest security, access control and data recovery standards.
The platform implements an end to end, fully automated screening and analysis process: Plate data from screening machines is uploaded to the cloud in real time. Proprietary big data analysis algorithms analyze millions of images and billions of data points to identify compounds that are active as translation modulators of the target protein.