When talking about societal trends, the things that cause massive shifts are exponential in nature. These things are catastrophic. They are like a swelling volcano at the outskirts of your city. You can look at it from day to day, reassuring yourself that nothing will happen, but deep down in your feet, you can feel an almost unconscious rumbling; deep down, you know better.
A picture is worth a thousand words and the pictures that we use to describe these exponential shifts are called semilog graphs. This is because one axis is essentially squished so that we can keep all of the data within the frame of the picture. In this article, I will describe some of these societal trends as curves, and the catastrophic curve surfer, Michael Fischbach.
As previously stated, we are getting worse and worse at producing antibiotics while nature is getting better at evolving around our defences. This depressing trend is entrenched within Eroom's law, which predicts a doubling of the cost to produce a new medicine, every nine years.
Above we see Eroom's law plotted onto a semilog graph; in 1952 a billion dollars would get about 80 drugs onto the market, in 2010, the same money only delivered 2 fifths of a new drug. I have made a claim, that this effect is driving antibiotics out of the pharmaceutical developmental pipeline. And as nature evolves around our existing defences, plague will spread again.
This evolutionary effect should be granted its own named curve. But while researching this article I could not find it. It would describe the exponential effect of how nature defeats our small molecule antibiotics over time. As a working antibiotic becomes more popular, it is prescribed to more people, and there is an exponential increase in its usage. As it is exposed to more bodies or ecosystems, in close quarters, it gives bacteria an opportunity to evolve. As bacteria innovate and share their genetic information through lateral viral transfer or through plasmids, our antibiotics are made obsolete. Let's call this the "Harvard Agar" graph since the video below so viscerally presents the phenomenon:
The previously mentioned, curves are working against our society but there are positive trends which can also be described with semilog graphs. I will list some of them now.
From a technical perspective, we are getting much much better at prospecting the bacterial ecosystem to find new antibiotics. New technologies that could only have been dreamed of by the original antibacterial prospectors are coming online. We are building super tools.
An example of such a tool is a computer. Computers are becoming much more powerful and cheaper every year. This effect, loosely described as Moore's law has prescriptively predicted a doubling of hardware capability every 1 to 2 years since 1971.
The equipment required to read a genome from a batch of cells (genetic sequencing) has been improving even faster than Moore's law. This can be seen in the diagram below:
The above graph compares Moore's law with Carlson's law. Both of these exponential trends can provide positive catastrophes. Moore's law predicts faster computing hardware, while Carlson's law, the curve with the sudden pitch downward, describes the falling cost to sequence a genome. These are two more exponential curves which are changing the fabric of our society.
Thousands of different bacterial genomes have been sequenced, and their data is available for those who want to look at them.
Michael Fischbach, of the University of California, has taken on this challenge. Instead of just looking, he has trained a machine learning algorithm to read and automatically identify bacterial genes which can produce antibiotics. He has taken this program and unleashed it within an ever growing database of bacterial genomes.
This approach takes advantage of another curve, for which I can not find a diagram or a name. It is the curve for algorithmic improvements. The underlying ways to ask a computer to perform complicated calculations has been improving much faster than Moore's law. This is especially true for the algorithms used by our machine learning approaches since they are heavily reliant on standard numerical methods:
Everyone knows Moore’s Law – a prediction made in 1965 by Intel co-founder Gordon Moore that the density of transistors in integrated circuits would continue to double every 1 to 2 years. (…) Even more remarkable – and even less widely understood – is that in many areas, performance gains due to improvements in algorithms have vastly exceeded even the dramatic performance gains due to increased processor speed.
The algorithms that we use today for speech recognition, for natural language translation, for chess playing, for logistics planning, have evolved remarkably in the past decade. It’s difficult to quantify the improvement, though, because it is as much in the realm of quality as of execution time.
In the field of numerical algorithms, however, the improvement can be quantified. Here is just one example, provided by Professor Martin Grötschel of Konrad-Zuse-Zentrum für Informationstechnik Berlin. Grötschel, an expert in optimisation, observes that a benchmark production planning model solved using linear programming would have taken 82 years to solve in 1988, using the computers and the linear programming algorithms of the day. Fifteen years later – in 2003 – this same model could be solved in roughly 1 minute, an improvement by a factor of roughly 43 million. Of this, a factor of roughly 1,000 was due to increased processor speed, whereas a factor of roughly 43,000 was due to improvements in algorithms! Grötschel also cites an algorithmic improvement of roughly 30,000 for mixed integer programming between 1991 and 2008.1
For lack of a better phrase, let's call this radical trend in the improvement of algorithms, Grötschel's law.
Lastly, I would be remiss if I did not mention the data itself. Genomes are trending towards being free to sequence, but the data that they produce in aggregate is overwhelming in its scale and complexity. The good news is that everyone is choking on data, and there is a wonderful buzz phrase that ensnares the technology needed to muster it: Big Data. The amount of data being produced is outpacing Moore's law since the economic factors below Carlson's law (price of sequencing) is improving faster than Moore's law and it is Carlson's law that is largely driving the amount of genomic data.
A lot of curves have been listed, so I will summarise them here:
- Eroom's law: Society is getting exponentially worse at making drugs.
- Harvard Agar curve: Bacteria are getting exponentially better at defeating our existing antibiotics.
- Moore's law: Society is getting exponentially better at making computer hardware.
- Carlson's law: Society is getting exponentially better at reading genomes.
- Grötschel's law: The important computer algorithms are improving at an exponential rate.
- Big Data: Society is generating data at an exponentially increasing rate; data management tools are tracking this trend.
Now imagine smashing these societal trends together to form a devastating confluence. Above this, you will find our catastrophic curve surfer, Michael Fischbach, dominating.
Michael Fischbach isn't digging in the dirt looking for new bacteria. Instead, he sits in front of his computer, and uses powerful machine learning algorithms to scan the increasing array of available bacterial genomes. He is being funded by the rising fear of superbugs, he is even being funded by DARPA. His research is focused on human biota because they have evolved within us, not in some dark corner of the ecosystem.
His lab ran their program over the genomes of 2500 organisms known to live within the human body and have discovered over 14000 biosynthetic gene clusters. With some more detective work, they discovered a new antibiotic called lactocillin, which is produced by a bacteria that is commonly found within human vaginas.2
Lactocillin has a targeted effect against Staphylococcus aureus, which is the cause of many types of bacterial vaginosis and other diseases typically called staph infections. You should know that at the time of this writing more than 10% of bloodstream Staphylococcus aureus infections in 15 European countries are caused by methicillin-resistant strains (MRSA)3. Superbugs.
If that wasn't enough for you, Michael Fischbach is participating in a revolution. There has been a holistic shift in biology lately, where animals are being seen as host-microbe ecosystems, rather than distinct and unique entities. In evolutionary theory, the hardline survival-of-the-fittest way of thinking is losing ground to the philosophy of symbiosis.
In a previous article, I wrote about how difficult it was to isolate a bacteria for study. But you should know that Michael Fischbach is more interested in communities of bacteria: culture, rather than how one may work in isolation.
This is because upon looking at the genomes of various bacteria he discovered that a large percentage of their genetic information was set aside to produce small antibiotic molecules. This genetic information was contiguous on the genome (which made it easier for his machine learning algorithm to find) and set in such a way that it would be easy to transfer between bacterial members of the community. Like open-source 3D gun printing blueprints posted on the pirate bay. Yet, these genes were not being expressed while the bacteria was studied in isolation. Why would a bacteria dedicate so much of its genome for the building of a complex small molecule and then not express it? Why would it be so willing to share its weaponized blueprints with other bacteria that it should be competing against?
It has been shown that the communities of bacteria living within us produce a-wash-of molecules that can be found in our blood, in concentrations that you would see if you were taking a drug prescribed by your doctor. These bacteria self-police; they get rid of bad guy germs that aren't playing well within their bacterial community and are pathogenic to the host. You are a walking pharmacy, you have your own FDA and you are a drug user.
Michael's team is going to identify the top 100 molecules that we get for free from our gut culture, then identify which specific gut bacteria produce each of these molecules. After that, he's going to sequence these bacteria to find their biosynthetic genetic information and use this information to train up his machine learning algorithm. With his updated algorithm he will be able to automatically find other bacteria that produce similar molecules.
Michael Fischbach is looking for things that are hidden in plain sight; in and on, the human body.
To understand the importance of this, think about how some pharmaceutical researchers spend their entire careers looking for effective medications, and never get one past the regulatory approval process. Now imagine that you told one of these frustrated researchers you could give them a sneak peek into another dimension full of humans with their own medications, most of which haven't been heard of here. These medicines have been used by this other society for a very long time, and if he would like he could analyse their chemistries, patent them as his own invention and monetize them as his own set of discoveries.
Better yet, why don't we give this researcher a search engine where he can scan more than one civilization at a time, looking for interesting pre-discovered and pre-used small molecules. This is what Michael Fischbach is building.
He is making an end run around the FDA, by using something that already has. The microbiome didn't ask for governmental permission to invent, produce and deliver drugs.
Once he understands how to build synthetic, custom, culture reactors, we can prescribe them to people suffering from a low metabolism, chrons disease, people who are low in iron, or people who are immuno-compromised. You could even prescribe something to elite athletes; performance enhancing poo. The list is endless.