What you need to know:
- Using notes from doctors, AI has further proved to be effective in predicting a patient’s rate of survival and life expectancy with an 80 per cent accuracy.
- In the revolutionary medical breakthrough displayed by different systems, scientists developed a drug that could potentially cure a type of liver cancer known as hepatocellular carcinoma.
Scientists’ efforts to use artificial intelligence to tackle deadly diseases could come to fruition after the technology developed treatment for a belligerent cancer in just 30 days.
Using notes from doctors, AI has further proved to be effective in predicting a patient’s rate of survival and life expectancy with an 80 per cent accuracy.
In the revolutionary medical breakthrough displayed by different systems, scientists developed a drug that could potentially cure a type of liver cancer known as hepatocellular carcinoma (HCC).
The HCC drug was created using Pharma, an AI platform for discovering treatment, through collaboration between Insilico Medicine (a biotechnology company based in Hong Kong and New York) and a team of researchers from the University of Toronto, Canada.
The AI system did not only develop the HCC drug, but it also discovered a treatment pathway that was initially unknown.
It went on to design what the medics termed as a ‘novel hit molecule’, which can bind to the set target.
The model was found to be highly accurate in predicting a patient’s life expectancy at 80 per cent, as established by scientists at the B.C Cancer and University of British Columbia.
According to the experts, it can also be applied to predict treatment effects, uncover relationships as well as patterns of deadly diseases.
Other than merely being used for texts and image generation, the technology could just become a modern weapon for combating terminal illnesses, thanks to its unmatched capacity to analyse enormous data amounts.
The team of scientists used a database of protein structure powered by the AI system known as AlphaFold to form and arrange a potential HCC cure.
“Our generative algorithms using AI succeeded in designing a target’s potent inhibitors with a structure derived from the AlphaFold, when advancements enabled by AI in language and art fascinated the world,” said a presser by Insilico Medicine director and founder Alex Zhavoronkov.
The impressive results were realised after blending just seven compounds and were achieved in merely 30 days.
The scientists also discovered a hit molecule that was more potent in their second phase of generating compounds using AI technology.
However, further clinical trials will still be required to be conducted on potential treatments, the medics advised.
“AlphaFold has pioneered a new scientific technique, which can be used to predict all the protein structures in the body of a human,” explained the researchers.
“We discovered a great opportunity to collect the structures and use them in our AI platform, which is end-to-end so as produce unique cures for tackling illnesses, which have a high rate of unmet need. The study is a vital first step towards that direction.”
Regarding the prediction on life expectancy, the researchers used a branch of AI which understands complex language of humans also known as Natural Language Processing (NLP).
They used the model to analyse notes made by an oncologist during the first consultation visit by cancer patients.
The NLP identified unique characteristics of each patient and predicted with an 80 per cent accuracy their survival rate at six months, 36 and 60 months.
“The AI has similar capability to humans of reading consultation documents.The notes comprise several details such as underlying health conditions of patients, their age, history of their families, cancer strain and past use of substances. It combines all these in order to provide outcomes of a patient’s complete picture.”
Currently, medics rely only on a handful genetic factors like tissue type and cancer site to calculate and categorise rates of survival of cancer patients by considering past events.
The AI technique is more efficient since it has the ability of picking unique clues in the first consultation document of a patient in order to generate an assessment that is more refined.
To test and train the AI module, the researchers collected data from 47,625 patients based at six centers in British Columbia. Apart from being highly scalable, neural NLP systems are portable and they do not need data sets that are structured.
This makes it possible to train them fast and improve a new region’s performance by applying local data.