Much like a bicycle leverages mechanical advantage to propel a person much faster and further than on his or her own, automation machinery is amplifying both the speed and accuracy with which biological research can be conducted. Machines are precise and do not fatigue, unlike their human counterparts. For this reason, machines are slowly being adopted for laborious, repetitive, and tedious laboratory tasks.
There are two ways of thinking about machines in the modern scientific community: physical, mechanical machines and laboratories, institutes, and facilities to which experiments can be outsourced. The former of these are often tailored to a single specific task. Tasks can include phenotypic screening, liquid transfer, and even laboratory organism maintenance. Mechanical machines are perfect for performing repetitive, tedious lab work inside the laboratory. Many of these machines offer sensors that can detect and measure traits that would be otherwise immeasurable by hand. When utilized properly, mechanical machines can greatly increase precision, efficiency, and reproducibility while reducing overall workload. Mechanical machines, while massively advantageous when working properly, often lack the ability to sense their surroundings or steps in their mechanical processes. Therefore, it is nearly impossible for these machines to adapt to conditional changes or for certain segments of the machine to provide feedback to other segments, either to adjust processes or alert their human keepers to an issue.
Great strides must be made in the design and engineering of experimental machinery so as to reduce innate variability of experimental setup and screening. Human’s, sentient beings, can sense and adjust to errors made mid-experiment. If too much of a solution is added to a reaction, a person can adjust the rest of the protocol on the fly or, at worst, begin the reaction over again. If the same error is made by a machine, it may go unnoticed until the output data are collected. Even worse, it may go completely unnoticed, presenting false data that are interpreted as correct upon analysis. Likewise, if the error is due to the improper functioning of a device, such as a pipette in the above example, a human is more likely to notice the error in the first place as we are equipped with sensors, i.e. eyes, that can detect that the volume transferred does not match the volume desired. Such sensors and feedback networks are currently not available on many of the mechanical machines being utilized in experiments.
The second way to think about machine automation in science is through the outsourcing of experiments to external research labs, institutions, or facilities. These organizations will take, as input, reagents or experimental parameters and return the output reagents or data back to the experimenter. These organizations can be thought of as automation machines because they take input and serve output from and to experimenters without any further involvement or work from the experimenter. The most common example of this process is that of DNA sequencing. While DNA sequencing has become a routine task in many modern laboratories, only a small minority of labs own and operate their own sequencing machine, either due to financial cost or lacking an operator with the requisite expertise and experience. Instead, most laboratories outsource this process to external organizations that specialize in DNA sequencing.
Perhaps the greatest advantage of the utilization of automation machinery is the ability
to pipeline experimental segments into one another. For example, laboratories can now outsource the creation of a specific reagent, the utilization of the reagent in a specific experiment, and the measurement of the outcome of the experiment all without conducting any work in their own laboratory. While the jury is still out on whether this pipelining technique will ever lead to any “virtual laboratories” where investigators simply dream up experiments, outsource their execution, then analyze the data, it is important to note that the modularization of experiments will almost certainly lead to increased efficiency in much of modern science. Instead of spending money on training and technicians to run repetitive short-term experiments, researchers now have the capacity to outsource these experiments to either mechanical machinery in their own lab or external research organizations for a fixed, per-experiment cost.
Automation can increase the productivity of a single individual by orders of magnitude. Machines do not tire or vary innately in their performance of a task; however, machines can break, suffer inaccuracy or variability in their measurements, and not detect faults when they may be blindingly apparent to a human. For these reasons, great care must be taken in protocol derivation and maintenance scheduling when utilizing any form of automation equipment.
When utilizing machinery in the design and/or execution of an experiment, the most important variable to consider is the trustworthiness of the data at each individual step. If unnoticed, systematic biases in data collection resulting from measurement error of machinery can lead researchers astray. Additionally, variability between experimental runs must be considered. One sign of data trustworthiness is reproducibility of the data. If the same experiment is replicated under the same conditions, the data resulting from each round of the experiment should be, at best, identical or, at worst, comparable. Although automation machinery has the incredible upside potential to streamline and parallelize experimental workflows, scientists must be careful to thoroughly validate results, both data and reagents.