The group aims to bring together Turing Fellows who have a research interest in genomics, proteomic, transcriptomic, metabolomic technologies to share best practice in experimental design and develop new methodologies to interpret these data through the application of machine learning, AI and novel computational and mathematical methods.
Explaining the science
Modern omic methods enable the generation of large amount of biological data with ever decreasing cost from a single experiment. However, badly designed experiments can result in worthless data. This interest group aims to ensure that these challenges are resolved before the first experiment has begun.
With the advent of whole-genome sequencing, single-cell sequencing and the advances in mass spectrometry methods, good experimental design and data analysis methods are more important than ever. Our objective is, therefore, to share best practice at the interface between data sciences and experimental biologists to ensure reproducible science and improve research outcomes. This meets the Turing goal to "advance world-class research and apply it to real-world problems."
What are the key challenges in designing a large scale omic experiment?
Challenges: Experiments are easily confounded by other variables, and poor experimental design may prevent answering the key research questions.
Example output: Publication of best practice methods.
How do we interpret machine learning outputs of biological data?
Challenges: Machine learning is incredible powerful; however, in the biomedical contact the results of model may be complex to interpret and implement the results.
Outcome: Strategies for design and interpreting machine learning strategies for real world impact.