Artificial Intelligence to Help Human Researchers Cure Ageing

Source: Deep Learning on Medium

The idea of curing all human diseases has recently become the egalitarian pastime of contemporary tech billionaires. Yes, you sort of get it, someone like Bill Gates and Facebook’s Zuck commonly known as Mark Zuckerberg.

The 64-year-old Microsoft founder is notable for his crusade dedicated to fighting against malaria disease focusing in third world countries like Africa. While the Facebook CEO has been rumoured to having the goal of curing all kinds of diseases by the end of the century.

He really zucked, at least in a useful way. Not bad, it’s for the betterment of the human species after all. And who cares who’s doing what when it’s for the common good ¯\_(ツ)_/¯

But how about the disease that nobody is talking about called ageing?

We know it’s there and most of us simply don’t care and just accept this awful degeneration that is happening inside and manifested outside as part of growing up, the peak of our biological maturity. Duh, whatever!

It’s like watching your house burning bigger and bigger as you feel the heat, and all you could do is nothing but behold until it all turned to dust. What the literal heck?!

Ageing is humanity’s dark fate

It’s what the society has ingrained into each one of us that we must embrace with graceful acceptance, and nobody seemed to challenge the status quo, until in recent years. Ageing is a tragic destiny we all must confront eventually if we do nothing to alter this existential dark fate.

A group of longevity scientists, however, have long taken this idea seriously by wanting to classify ageing as a disease. Classifying this kind of natural anomaly should be the first step to finding its cure.

Life extension medicine and intervention are some of the main highlights of the company Insilico Medicine, a now Hong Kong-based startup which makes use of next-generation AI for drug discovery and longevity exploration.

The company uses a deep learning model called GAN and RL, short for Generative Adversarial Network and Reinforcement Learning, a similar technique used in DeepFake which generates photorealistic images of imaginary people that are indistinguishable from real ones.

Instead of generating synthetic images, its artificial neural networks were trained using several existing datasets to generate unique drug properties targeted at currently incurable diseases. Recent achievement demonstrated how its AI system is capable of designing a new drug candidate within weeks as opposed to conventional drug development that usually lasts several years with billions of research funding.

Deep learning for life extension sciences

In line with its unrelenting effort to advance the field of medicine using AI, its glorified deep learning model is also being applied to identify biomarkers of ageing. These biomarkers are essential in predicting a patient’s true ‘biological age’ in contrast to a chronological timeline which is based on the number of years and days that a person has lived thus far.

One of the greatest challenges the longevity field is facing today is the identification of accurate ageing biomarkers. Researchers lack this crucial data which causes a stall in the development of therapies for certain age-related diseases, and ultimately antiageing therapeutics.

This so-called ‘deep ageing clock’, sort of biomarkers that the AI identified can be used to forecast a patient’s underlying health condition that may soon arise when left untreated allowing for advanced intervention before the onset of any age-related diseases.

A lot of hidden variables involved as to why some group of people age faster than the others, and this also means lots of work to be done. Aside from genetics, this could be the result of lifestyle, diet, pollution, and a myriad of variables too vastly complicated for human researchers to unravel within a constrained breadth of understanding and insufficient data.

Luckily, this is something that applied deep learning for medicine naturally excels at. Not because it involves supercomputing capability, but because it was designed from the get-go to do exactly what it’s supposed to accomplish. That is to conceptualize things that aren’t yet there which we necessitate to be there without the need for enormous loads of datasets.

This type of advanced neural network algorithm is what set apart from the rest of existing machine learning models like pattern and image recognition, etc, which heavily rely on crunching massive databases and low-level neural programming.