Why Does AIQC Exist?
Over the past 4 years, I worked with the top 5 pharmaceutical companies to analyze national biobanks, such as the UK Biobank and Genomics Medicine Ireland, for the genomic-drivers of complex diseases.
In the face of such challenging & important problems, I was shocked that big pharma’s primary form of analysis was the basic statistical test known as an association study, which dates back to the Victorian era. I kept expecting someone to say, “Okay, now is the time for us to start using deep learning,” but it never happened. If the researchers at the most well-financed companies in the world weren’t equipped to take advantage of AI, then how would it ever be possible for non-profit scientists?
Deep learning has the power to accelerate the rate of scientific discovery by acting as a torch that reveals the laws of nature through data-driven pattern recognition. When it comes to global crises like combatting pandemics and reversing climate catastrophe, the human race is at a point where it needs to make major scientific advances over a short period of time in order to survive. So let’s empower our smartest people with the best analytical tools we have.
1. Accelerate science by making deep learning accessible.
Reduce the amount of programming and data science know-how required to perform deep learning. This unattainable skillset trifecta causes machine learning to be underutilized in science. What would Newton & Einstein have discovered with the power of deep learning?
Provide field-specific deep learning solutions for research in the form of: pipelines for preprocessing scientific file formats, pre-trained models for transfer learning, and visualizations of predictions.
2. Bring the scientific method to data science.
Make machine learning less of a black box by implementing “Quality control (QC)” protocols comprised of best practice validation rules.
Reproducibly record not only the machine learning experiments, but also the lineage for preparing data. This is important for combatting bias during the data gathering and evaluation phases.
3. Break down walled gardens to keep science open.
This toolset provides research teams a standardized method for ML-based evidence, as opposed to each research team cobbling together their own approach. An AIQC file should be submitted alongside publications and model zoo entries as a proof.
The majority of research doesn’t happen in the cloud, it’s performed on the personal computers of individuals. We empower the non-cloud researchers: the academic/ institute HPCers, the remote server SSH’ers, and everyday laptop warriors.
If the entire scientific community does not have access to the toolset used to conduct the experiment, then it is not reproducible.