Wednesday, February 4, 2015

Why Lab Automation is NOT the Future of Science but Networking and Remote Access IS


So PCR is one of the most simple and basic molecular biology experiments to do.

Let's try and envision what an automated PCR system would entail...

1. Every primer, template, enzyme and buffer would need to be stored and labelled in such a way that a robot could access them. This is difficult given that most of these should and need to be stored in at least -20C for long-term storage. So we need a robot accessible freezer system with each individual tube barcoded or placed in a known spot. This is extremely cost and space prohibitive. Also, creating the database of every component would take effort.

2. You need a robot that can pipette and move liquids that are not frozen so you need it outside of the freezer storage system.

3. You need a robot to move the tubes into the PCR machine or some setup where the PCR machine is integrated into #2.

4. You need someone to tell the robot what primers, DNA, enzyme, buffers and PCR protocol to use.

After you spend probably $5 million+ for this system. It would still take someone probably the same amount of time to use the program to setup the robots that it would take someone to setup 8 PCR reactions, that is, less than 5 minutes. In the long run it would probably take much much more time in maintenance and troubleshooting issues with the robot hardware and software.

The majority of researchers in the majority of labs perform probably around 8 or less PCR reactions per experiment. This is important because Lab Automation will only succeed when time and monetary investment is less than the time and monetary investment of doing it manually and currently it is a far way off.

For millions of dollars in investment one expects a couple orders of magnitude in time saved. This system would not even help if it is scaled up by say, doing 96 PCR reactions. If each reaction uses an individual template or primers at about 20 seconds per manual primer addition per well(everything else can be done in about 2 minutes with a multichannel pipette and mastermix) you are looking at about 32 minutes. To manual enter the names of each individual primer into a program for 96 different wells would take at best 5-10 seconds per primer. So even for 96 well plates one is only saving researchers half their time and again this doesn't include troubleshooting if say the robot has trouble removing a tube because condensation built-up and froze on the storage box. If someone used a robot everyday and saved themselves 15 minutes at the end of the year they would have saved themselves 91 hours for millions of dollars. How many hours a year are worth $5 million? The average Science researcher(grad students and post docs) probably work >2,000 hours a year and are paid around < $45k. For $5 million, one could hire ~111 people to do work for a year and there is no service contract required for people, they are pretty robust and don't break down very often.

The idea that everything can be automated is not true and especially not true for Science and probably will not be for a long time. There are very niche uses for automation robots in Science but these are very unique circumstances and not an everyday type of thing.

People who think lab automation is the Future of Science have not spent much time in a lab.In fact I have yet to meet someone who has spent alot of time in a lab who believes automation will help Science. I am sure there are people out there but it is definitely a minority.

So what does Science need then?

Science really needs Remote Access and Remote Control. Say on my way to work in the morning I figure I need turn on a couple of incubators or change the temperature in some because I had a culture expressing proteins. Maybe I wanted to equilibrate my H(F)PLC column? What if I could do this from my cell phone? The thing that needs to be optimized in Science is not the "work time" it is the "down time". When I need to wait 30 minutes for my H(F)PLC column to equilibrate and cannot use it or when I need to wait for a centrifuge to cool down or an incubator to heat up or many of the other things. This is where time is wasted. Transferring data files between computers or restarting a experiment on a piece of hardware remotely this would drastically help Scientific productivity.

Networking labs and providing things like remote access costs a fraction of automation. We are talking in the $10ks to $100ks per lab.

I really tried to have this instituted at our NMR facility at UChicago because I had to head down to the NMR room, about a 5 minute walk, anytime I wanted to start an experiment or access data, a computer not connected to the outside internet was the interface to the Spectrometer. I would have saved myself hours of down time because I could have spent 5 minutes using remote access to start another experiment from home or restart an experiment with a different parameter, time wasted because I wasn't on campus or was far enough away that it impeded me going to the NMR room at the moment.

Labs are for sure still stuck in the 1970s but not in an automation sort of way, in an access sort of way.

It's 2015, I work at NASA(we are supposed to be high tech(yeah right I know NASA is still stuck in the 1970s)) and I can't even access the computer that was purchased brand new last year that is in the other room without walking there and sitting at the console. That is a problem...

I even started to work on this problem a little bit by making Arduinos controllable through a webpage so they could switch relays or run equipment (http://doitourselfscience.blogspot.com/2014/03/sending-and-receiving-data-from-arduino.html). After the basic setup though I figured I would rather spend time doing research then building such as system. Foo on me.