As mentioned in the previous post, some researchers are starting to tackle the problems of low efficiency, poor repeatability, and safety in synthetic experiments in an automated and intelligent way. In order to give you a deeper understanding, we decided to start a series to introduce these scientists and their achievements one by one, and we will continue to track their latest work in the future. I hope to make a lost contribution to the laboratory model that experimenters will change their minds and embrace automation, intelligence and IoT. The content is wonderful, follow us, will not be lost ( ̄▽ ̄)ノ
The first person to introduce is Professor Alan Aspuru-Guzik from the University of Toronto.
Alan Aspuru-Guzik is a professor of chemistry and computer science at the University of Toronto, and has long been involved in quantum information, chemistry, and the fusion of machine learning and chemistry. He was a pioneer in the development of algorithms and the implementation of experiments that specifically used quantum computers for chemical systems and quantum simulators. Recently, however, Alán has become more interested in automated chemistry labs that accelerate scientific discoveries, and has produced remarkable results in both academia and industry.
01
Goal & Solution
Academically, for Alan Aspuru-Guzik's "AI for Discovery and Self-Driving Labs" research group, the goal is to reduce the time and money required to discover new functional materials or optimize known functional materials by a factor of 10, from an estimated $10 million and 10-year development time to $1 million and 1 year. The solution they propose to address this challenge is to develop self-driven labs that combine artificial intelligence with automated robotic platforms to autonomously discover new materials. Creating a fully autonomous autonomous driving lab is a multidisciplinary task, and the team explored a number of areas, including the use of artificial intelligence to design optimization algorithms for controlling and designing experiments, robotic systems for performing these experiments, and automatic characterization methods for analyzing results, which form a closed loop in the Alan Aspuru-Guzik laboratory in an intelligent, IoT-linked model, minimizing the low-end artificial components in the experimental process, and focusing on innovation and design. Of course, this laboratory model also demonstrates the productivity of its closed-loop autonomous discovery in several jobs: (1) generating approximately 40 organic semiconductor molecules for laser equipment in a single run (2) greatly optimizing the composition and processing parameters of thin film materials over a wide space (SCIENCE ADVANCE, 13 May 2020, Vol 6, Issue 20, DOI: 10.1126/sciadv.aaz8867).
02
Automated laboratory composition
Let's briefly look at the four components of this laboratory cycle pattern – virtual screening, robotic platform synthesis, automated characterization, and "brain" control software – centos.
In Alan's lab, virtual screening for high-throughput calculations based on quantum chemistry and computational chemistry is their specialty. Through computational simulation and machine learning, the appropriate range of molecules is handed over to the Chemspeed robotic platform in the laboratory for high-throughput sample preparation, synthesis, process development, formulation, application or testing, accelerating product development.
Chemspeed device
Product synthesis is followed by their automated characterization platform, in addition to communication and product delivery between multiple traditionally characterized instruments. Computer vision, which is shining in the fields of face recognition and autonomous driving, is also used as a new research direction to enhance visual perception when conducting chemical and materials science experiments, such as real-time observation of liquid types, phases of materials (e.g., liquids, solids, foams, suspensions), types of containers, and characteristics of containers (transparent and opaque) during experiments.
Use computer vision to put "eyes" on the lab
ChemOS software acts as a brain, managing data collection, experimental procedures, and associated robotic devices. It also supports remote control of devices, which allows ChemOS to run across different labs, even labs located in different institutions. ChemOS helps different labs share complementary areas of expertise, accelerating the discovery and innovation process. One such "shared" laboratory was established between Vancouver, Canada, and Cambridge, USA, where ChemOS calibrated a robotic sampling sequence for direct injection hplc analysis. The process ran completely autonomously for several days without supervision, accumulating a total of 1100 experiments designed by ChemOS.
03
Outside the lab...
In addition to his lab work, he is also the editor-in-chief of Digital Discovery, a newly created Open Access journal of the Royal Society of Chemistry in August 2021, which caters positively to the trend of high degree of automation, data-driven integration with chemistry, materials science, and biotechnology, with the goal of capturing top research in this intersectional field, as well as topics related to machine learning, high-throughput computing, and experimental screening, to accelerate the process of scientific discovery.
Of course, Alan Aspuru-Guzik is not only entertaining himself in his own laboratory, he is also actively pioneering in the industry. Aspuru-Guzik, which jointly launched Kebotix at Harvard University in 2017, is preparing machine learning and robotic automation to revolutionize materials science in the coming years, as the company's slogan puts it: At the push of a button, advanced chemicals and materials can be discovered and created at a faster pace. Kebotix, which recently raised $5 million in funding and was led by One Way Ventures, stepped out of stealth mode. Investors also include Baidu Ventures, Bridge Capital Partner in Boston, Embassy Ventures in Los Angeles, Propagator Ventures in Norway and WorldQuant Ventures in New York. Recently, the Kebotix lab identified a handful of molecules that can be used as electrochromic glass materials out of 7 million molecules sold on the market, and AI has also designed hundreds of additional new electrochromic molecules, each of which meets the target properties. This work proves that Kebotix is discovering a new type of material at an extremely fast pace to prove that it is fulfilling the mission of Kebotix when it was founded.
KEBOTIX: https://www.kebotix.com/
In addition, he coordinated the formation of a consortium called the Acceleration Consortium, a new global collaboration between academia, industry and government, based at the University of Toronto (U of T). The Acceleration Consortium has three interrelated goals: (1) to transform technological innovation: to advance the design of the Materials Acceleration Platform (MAPS) to accelerate the discovery of new materials and make fundamental breakthroughs in THE fields of AI, robotics, computing, and materials science; and (2) to build an innovation ecosystem: to build academic institutions committed to materials innovation. A global network of tech companies and entrepreneurs; (3) Developing a highly skilled workforce: developing a nationwide training program for the next generation of researchers.
Alan Aspuru-Guzik's Lab Home: https://www.matter.toronto.edu/
04
Next issue is coming soon
After introducing Alan Aspuru-Guzik, a hexagonal all-rounder from Canada, in the next issue we will bring a Twitter celebrity from the United Kingdom. Rare not bald. Professional cat breeding. Lee Cronin, who side-researched.