-
Categories
-
Pharmaceutical Intermediates
-
Active Pharmaceutical Ingredients
-
Food Additives
- Industrial Coatings
- Agrochemicals
- Dyes and Pigments
- Surfactant
- Flavors and Fragrances
- Chemical Reagents
- Catalyst and Auxiliary
- Natural Products
- Inorganic Chemistry
-
Organic Chemistry
-
Biochemical Engineering
- Analytical Chemistry
-
Cosmetic Ingredient
- Water Treatment Chemical
-
Pharmaceutical Intermediates
Promotion
ECHEMI Mall
Wholesale
Weekly Price
Exhibition
News
-
Trade Service
Image: Justin Pritchard, assistant professor of biomedical engineering, holds a mock genetic sample
.
This sample, which is less than a tenth of an ounce (1-2 milliliters) in volume, represents the size of a "compressed" gene library, compared to about 3 ounces (80 milliliters) for traditional libraries
Source: elby Hochreither/Penn State University
In image compression, storing or sharing large files can be cumbersome, so a small amount of visual information is lost
.
This "attrition" largely preserves the image while greatly reducing file size and provides inspiration for new research directions in genomics, said Justin Pritchard, assistant professor of biomedical engineering
An interdisciplinary team of researchers led by Penn State has developed a method to "compress" extensive genetic databases to a more manageable size
.
They published their findings Feb.
"This idea of compression dramatically reduces the size of experiments and opens up possibilities for new experiments," Pritchard said, "which could unlock biological mysteries, such as why different genes and drugs work together differently.
, it allows us to solve very complex biological mysteries with simpler experiments
.
"
The researchers point to genome-scale CRISPR experiments, which contain data on the effects of thousands of genes tested in different human cell types
.
The effects when genes are turned off vary by cell type, so large numbers of cells are needed to understand the interactions between genes and phenotypes
To predict larger genome-scale effects from smaller "compressed" CRISPR libraries, the team used a custom algorithm based on a common machine learning technique known as random forests
.
This approach integrates the data provided by the researchers into a series of randomly generated decision trees that together produce predictions about the relationship between gene inactivation and cell growth
This performance can be achieved using only a small percentage -- about 1% -- of information from the original library
.
Finally, the Penn State team conducted new experiments in which they actually built these "lossy compression libraries" using synthetic biology techniques, validating the predictions in new experiments
"A genome-scale experiment will probe 18,000 genes," Pritchard said
.
"Using machine learning, we compressed the experimental size to 200 genes
The technology also opens up opportunities for other research
.
It showed transferability, meaning that although it was only trained on CRISPR data, it could obtain accurate prediction matches from a completely different dataset
"We're excited about the future of this research," Pritchard said
.
"We can use newer machine learning techniques to change the composition of these lossy compressed sets in real time for different experimental questions and conditions ranging from cancer biology to biopharmaceuticals
.
The method also helps us by answering questions about how the genome works and encodes cells The problem of growing information to improve basic science
.
"
article title
A pan-CRISPR analysis of mammalian cell specificity identifies ultra-compact sgRNA subsets for genome-scale experiments
