Will AI protein generation accelerate drug development?
Rather unexpectedly, the Chalmers University of Technology, a leading institution in Sweden, recently released that demonstrates artificial intelligence’s (AI) newfound capability to generate novel, functionally active .
“What we are now able to demonstrate offers fantastic potential for a number of future applications, such as faster and more cost-efficient development of protein-based drugs,” said Associate Professor Aleksej Zelezniak, lead researcher of the study.
The researchers shared that protein-based drugs are very common ─ the most commonly-prescribed being insulin for diabetic patients who require a self-administered injection each day to maintain their blood sugar levels and maintain a normal, healthy life. The greater majority of most effective, and subsequently most expensive, cancer treatments are also protein-based, and, appropriate to the world’s struggle of the day, so too are the antibody formulas that are being used to create the incredibly important COVID-19 vaccinations.
Developing Superior Methods
At the moment, protein engineering methods rely on the introduction of random mutations to existing protein sequences ─ like most experiments, really. The problem is, every time scientists introduce a new mutation to the sequence, the protein actively declines.
“Consequently, one must perform multiple rounds of very expensive and time-consuming experiments, screening millions of variants, to engineer proteins and enzymes that end up being significantly different from those found in nature,” said Zelezniak. “This engineering process is very slow, but now we have an AI-based method where we can go from computer design to working protein in just a few weeks.”
ProteinGAN’s AI-powered Synthesis
A new AI-based approach to protein engineering has been introduced to the world from a group effort between the researchers and external collaborators: ProteinGAN. This method takes a generative deep learning approach to protein synthesis.
Like all AI and machine learning (ML) based systems, ProteinGAN can be fed repositories of data from already studied proteins; it then analyses the data with sophisticated analysis tools and starts to create new ideas and subsequent proteins of its own. Simultaneously, other parts of the AI work out whether the synthetic proteins that it creates are fake or not ─ all of the created proteins go through the system on a loop until the AI can’t tell the natural and synthetic proteins apart anymore.
A Cost-efficient and Increasingly Sustainable Solution
Martin Engqvist, Assistant Professor in the Department of Biology and Biological Engineering, was part of the team that designed the experiments to test the AI-synthesised proteins. On the topic, he stated that “Accelerating the rate at which we engineer proteins is very important for driving down development costs for enzyme catalysts. This is the key to realising environmentally sustainable industrial processes and consumer products, and our AI model, as well as future models, will enable that. Our work is a vital contribution in that context.”
“This kind of work is only possible in the type of multidisciplinary environment that exists at our Division – at the interface of computer science and biology. We have perfect conditions to experimentally test the properties of these AI-designed proteins,” Zelezniak added.
Facebook Develops AI to Crackdown on Deepfakes
In light of the large tidal wave of increasingly believable deepfake images and videos that have been hitting the feeds of every major social media and news outlet in recent years, global organisations have started to consider the risk factor behind them. While the majority of deepfakes are created purely for amusement, their increasing sophistication is leading to a very simple question: What happens when a deepfake is produced not for amusement, but for malicious intent on a grander scale?
Yesterday, Facebook revealed that it was also concerned by that very question and that it had decided to take a stand against deepfakes. In partnership with Michigan State University, the social media giant presented “a research method of detecting and attributing deepfakes that relies on reverse engineering from a single AI-generated image to the generative model used to produce it.”
The promise is that Facebook’s method will facilitate deepfake detection and tracing in real-world settings, where the deepfake image itself is often the only information detectors have to work with.
Why Reverse Engineering?
Right now, researchers identify deepfakes through two primary methods: detection, which distinguishes between real and deepfake images, and image attribution, which identifies whether the image was generated using one of the AI’s training models. But generative photo techniques have advanced in scale and sophistication over the past few years, and the old strategies are no longer sufficient.
First, there are only so many images presented in AI training. If the deepfake was generated by an unknown, alternative model, even artificial intelligence won’t be able to spot it—at least, until now. Reverse engineering, common practice in machine learning (ML), can uncover unique patterns left by the generating model, regardless of whether it was included in the AI’s training set. This helps discover coordinated deepfake attacks or other instances in which multiple deepfakes come from the same source.
How It Works
Before we could use deep learning to generate images, criminals and other ill-intentioned actors had a limited amount of options. Cameras only had so many tools at their disposal, and most researchers could easily identify certain makes and models. But deep learning has ushered in an age of endless options, and as a result, it’s grown increasingly difficult to identify deepfakes.
To counteract this, Facebook ran deepfakes through a fingerprint estimation network (FEN) to estimate some of their details. Fingerprints are essentially patterns left on an image due to manufacturing imperfections, and they help identify where the image came from. By evaluating the fingerprint magnitude, repetition frequency, and symmetrical frequency, Facebook then applied those constraints to predict the model’s hyperparameters.
What are hyperparameters? If you imagine a generative model as a car, hyperparameters are similar to the engine components: certain properties that distinguish your fancy automobile from others on the market. ‘Our reverse engineering technique is somewhat like recognising [the engine] components of a car based on how it sounds’, Facebook explained, ‘even if this is a new car we’ve never heard of before’.
What Did They Find?
‘On standard benchmarks, we get state-of-the-art results’, said Facebook research lead Tal Hassner. Facebook added that the fingerprint estimation network (FEN) method can be used for not only model parsing, but detection and image attribution. While this research is the first of its kind, making it difficult to assess the results, the future looks promising.
Facebook’s AI will introduce model parsing for real-world applications, increasing our understanding of deepfake detection. As cybersecurity attacks proliferate, and generative AI falls into the hands of those who would do us harm, this method could help the ‘good guys’ stay one step ahead. As Hassner explained: ‘This is a cat-and-mouse game, and it continues to be a cat-and-mouse game’.