From the assembly line to the black light lab, the automation revolution upstream of the life sciences

Second, when local manufacturers break through

Strictly speaking, the domestic market is still in the early stage of development, so many different forms of products and multiple landing scenes, where to find a breakthrough? Three basic requirements must be met: first, a full understanding of the requirements and results of automation from the goal; Secondly, the technology can be realized, and the process is relatively fixed, meeting the premise of landing; Finally, the system/product is mature and reliable, and customers pay for it, forming a commercial closed loop.

Result-oriented: replicating experimental results and exploring new mechanisms

Results-oriented, we first need to know where the demand for laboratory automation is reflected. If you refer to industrial automation, the key is to reduce costs and improve efficiency, and the final result is a processed or assembled product.

In contrast, laboratory automation ultimately produces results that can be divided into two categories: first, the completion of repetitive actions to produce an accurate data, which is mainly for detection laboratories; The second is to obtain data through a designed experimental process, explore and understand new mechanisms, or screen target compounds, which is mainly for R&D laboratories.

The operation steps and processes involved in the above two processes are mainly for the repetition of the experiment. At present, the application of the detection scenario is relatively mature, and the research and development scenario often needs to use the experimental data to analyze the experimental steps and the specific details of the experiment, so as to optimize the experiment. From this point of view, laboratory automation is not only the operation of the action execution level, but also includes the sample flow, information flow and operation flow. In addition to the operation of the experimental instrument, it is also necessary to integrate the results of the experimental instrument or the data results of the experimental process.

The realization of the above functions generally needs to meet four requirements: AI engines, mobile platforms, multi-modal sensors, etc., to achieve a full range of data collection in the laboratory; Access third-party instruments through the central control system to optimize the working mode; The data of the whole process of the experiment were collected by visual sensors, mechanical sensors, etc., and the experimental process was continuously optimized by steps that were not normally observed or analyzed. AGV robots are introduced to ensure flexibility and connect physically more segmented "independent functional areas".

Landing premise: technology can be realized, to process

For the automated exploration of testing laboratories, its technical threshold is slightly lower, the current industry has relatively large-scale or mature applications, mainly for independent testing, clinical diagnosis of standard projects, and nucleic acid testing due to the epidemic "multiplied ten times", they generally have a large market size, but most of these areas have become competitive Red Sea.

For the exploration of research and development laboratories, entrepreneurs generally reflect that in fact, the above technology is not difficult to separate, difficult to integrate it, achieve full automation, and well cope with the "complexity" and "non-standard" of life science scenes - these are high flexibility and precision requirements, as well as the rise of demand in some emerging technology fields.

Typical new technologies include gene editing technology, IPS cell culture technology, etc., which will bring great changes to human health and life. In this process, automation and intelligence will greatly promote the development of the industry. Taking the gene sequencing industry as an example, automated sequencing technology throughput is now more than 10 million times higher than it was 20 years ago, and its development rate is far faster than Moore's Law in the semiconductor industry. Similarly, the premise of scale application in some emerging fields is also that the automation technology involved in many links has been broken through, so that things that were very difficult to do in the past can now be solved by processes and standardized means.

For example, Alphafold2's cracking of the protein molecular folding problem for amino acid sequence prediction, the progress of AI in crystal type prediction, and the scale application of mRNA therapy technology have all turned biological problems into computational problems. As well as the evolution of semiconductor technology, the computing power breakthrough has opened up the path of computing power to the research and development of new biological drugs. If these scattered technologies can be gradually connected, the entire mass trial and error behavior of biomedicine can be turned into repeated work by machines.

Taking the automated development of macromolecular drugs as an example, at present, deep learning has a significant effect on the improvement of protein structure prediction, which can accurately predict the rough shape of most proteins in three-dimensional space, and help biotechnologists identify and produce proteins, so as to make the research and development of innovative macromolecular drugs predictable and programmable, and improve the efficiency of the whole process of drug research and development. However, most of the subdivision directions are still in the early stage of exploration, and there are fewer actual landing cases.