What Can Cancer Research Learn From ‘The Origin Of Life’?

Despite enormous progress over the past century, cancer is still something that ends life. Why then should cancer researchers care about the polar opposite? Why should we care about the origins of life?

This was the challenge facing UCL’s Dr Nick Lane as he presented the ICR’s annual ‘Darwin Lecture’ last Thursday — 208 years (to the week) from Charles Darwin’s birth. With an enigmatic title of ‘The Origins of Life on Planet Earth’, Dr Lane had attracted teeming audiences in both Chelsea and Sutton. Both lecture theatres were standing room only. Cancer biologists, it seems, are rather interested in such big questions. 

Dr Lane started by describing the great ‘evolutionary scandal’ at the heart of biology. Why, he asked, do we not see eukaryotic traits in prokaryotes? Why did eukaryotic complexity not co-evolve in bacteria? The answer, he revealed, lies in bioenergetics. Dr Lane explained how eukaryotic life originated from a single ‘big bang’ event 1.5–2.0 billion years ago, when an archaeon engulfed a proteobacteria. This endosymbiosed aerobic bacteria enabled the host archaea to massively increase its energy production — providing the metabolic resources to develop further complexity. Eukaryotes had arisen not from a linear evolution of one cell type, but from a combination of different cells. Less of a ‘tree of life’ and more a ‘ring of life’. Empowered by their newfound energy, early eukaryotes eventually evolved into the pantheon of multicellular biology we see today. It was a humbling perspective of our cellular origins. 

I think it’s fair to say that ICR scientists rarely consider such topics. In many ways, ‘origin of life’ research is the ultimate blue sky endeavour — studied because it’s fascinating, not because it’s obviously useful. This research is somewhat removed from the ‘translational research’ the ICR aspires to. Yet the abundant audiences across both ICR sites demonstrated there is a real craving for such insight among translational scientists. There is something universally intriguing about our ancient common ancestors. 

Dr. Nick Lane discussing ‘The Origins of Life on Planet Earth’ at The ICR, London

Dr. Nick Lane discussing ‘The Origins of Life on Planet Earth’ at The ICR, London

Implications for cancer research 

So aside from being genuinely fascinating, how does the ‘origin of life’ relate to our day-to-day research of cancer? What can we learn from our cellular history?

From speaking to colleagues after the lecture, everyone took away something different. The genetically inclined were fascinated by how mitochondria lost 99% of their genes just so they could spend the saved ATP on something new. What evolutionary advantages might be afforded by gene loss frequently seen in cancer cells?

The metabolically enthused certainly enjoyed seeing bioenergetics at the centre of cellular evolution. Given metabolic changes are powerful enough to drive the entire evolution of eukaryotic life, maybe we shouldn’t be surprised by its frequent deregulation in cancer. Could cancer just be an intra-organism metabolic shift?

As someone interested in the tumour microenvironment, I was charmed to learn how the eukaryotes, with their great potential for complexity, arose not by serial evolution of a single cell, but by a symbiotic collaboration between two different cell types. In cancer, we often see healthy cells (such as fibroblasts, endothelial, and immune cells) coerced by cancer cells to provide tumours with resources cancer cells can’t make themselves. It’s fascinating to think our eukaryotic ancestors were formed by a similar collaboration of different cells. It implies the improved ‘fitness’ afforded by heterocellular collaboration is both ancient and universal.  

Whatever you took away from the lecture, I think it’s fair to say everyone left with a new view on our cellular origins and their consciousness substantially raised. I’m already looking forward to Darwin’s 209th.


This post originally appeared as a 'From Bench To Blog' blog for The Institute of Cancer Research.

Heterocellular Emergence

Humans are jam packed with different cell types. Throughout your body you'll find epithelial cells, mesenchymal cells, nerve cells, endothelial cells, and a bewildering diversity of immune cells

This complexity begs the question: why does biology bother? Can't we just have one cell type? What's the value in having so many different cells?

First, let's consider what makes each cell type different. For the most part, it's not DNA. The genome in your epithelial cells is largely identical to that in your fibroblasts. In contrast, the proteins expressed between different cell types are discrete. For example, epithelial cells express cell-cell adhesion proteins (enabling them to form an epithelium), whereas fibroblasts secrete extracellular matrix proteins (enabling them to organise tissue structure). Cell-specific proteomes enable cells to achieve specialised functions. 

Signalling Societies 

One consequence of cell-specific protein expression is that each cell type contains distinct signalling proteins. This enables each cell type to process signals completely differently. For example, the same protein cue can tell one cell type to grow, yet tell another cell to die. This cell-specific processing is called 'homocellular signalling'.

Although homocellular signalling is what most cell biologists study, it's a simplified view. In reality, all cells sit alongside other cell types in tissues. 

When multiple cell types are combined, they can communicate with one another. This is called 'heterocellular signalling'. 

Heterocellular signalling enables specialised cells to exchange information and expand their collective signal processing capacity. When viewed from this heterocellular perspective, tissues resemble diverse cellular societies — and despite right-wing rhetoric, societies benefit from diversity.

Specialisation and Exchange

In socioeconomic theory, there is a basic concept called 'specialisation and exchange'. The idea states that it's more productive for workers in a society to specialise in skills and then exchange these services  than to try and do everything yourself. For example, it would be extremely inefficient for everyone in a society to train to be a doctor, a computer programmer, and an actor. We'd end up with worse doctors, worse programmers, and Hollyoaks as the pinnacle of thespian achievement. Instead, some people specialise in medicine, some become expert programmers, and others go to acting school. When the doctor needs some software, they'll buy it from the programmer. When the programmer gets ill, they'll visit the doctor. When the doctor has time off (should such a thing happen), they can watch Daniel Day-Lewis

By specialising and then exchanging skills, all parties get a better service and the whole society is more productive.

'Specialisation and exchange' is also found throughout biology. In tissues, homocellular signalling allows cells to specialise. Heterocellular signalling then allows cells to exchange information. This enables cells to combine forces to achieve phenotypes that no cell type can accomplish in isolation. 

For example, consider something you use every day: your gut. Intestinal epithelial cells have specialised homocellular signalling that allow them to form an epithelium. This creates a barrier between the food you ate last night and your blood. However, if epithelial cells are combined with myeloid cells (antigen recognition) and lymphoid cells (antigen attacking), they can form a tissue epithelium with adaptive immunosurveillance. The diversity of multiple cell types creates a smart barrier that can simultaneously detect and kill pathogens. Such a complex phenotype can only be achieved when diverse cells work together no one cell type can do it all.  

Such collaborative behaviour is common in biology. Adaptive immunity, sight, digestion, and homeostasis are all complex phenotypes that require multiple cell types. In fact, pretty much every tissue phenotype can be said to 'supervene' upon heterocellular signalling.

Despite its ubiquity, the process by which multiple cell types collaborate to achieve complex phenotypes doesn't actually have a formal name. Biologists frequently discuss the products of such processes, but there is no turn-of-phrase to capture cellular specialisation and exchange. 

In the absence of such language, I've taken the liberty of creating a term. I call it 'heterocellular emergence'. 

Here's the 'dictionary' definition:

Heterocellular Emergence | ˌhɛt(ə)rəʊˈsɛljʊlə ɪˈməːdʒ(ə)ns | noun

  1. a process whereby complex tissue-level phenotypes are achieved through interactions between different cell types.

The Heterocellular Emergence of Cancer

Heterocellular emergence is found throughout metazoan life — and like many processes in healthy tissues, heterocellular emergence is also seen in cancer.

Just like healthy intestinal tissue, colorectal cancer (CRC) tumours also contain epithelial, mesenchymal, lymphoid, and myeloid cells. And, like healthy tissues, CRC tumours use these different cells to achieve complex phenotypes. The difference with cancer is that these phenotypes (such as immune evasion and metastasis) can kill, rather than aid, the host organism. 

In a new article for Trends in Cancer entitled 'The Heterocellular Emergence of Colorectal Cancer' I argue that cancer is an emergent heterocellular phenotype. 

It's half review, half an attempt to unify anecdotal tumour microenvironment observations into an emergent theory of malignancy. If you're interested in how different cell types collaborate in cancer, check it out. 

Escape From Reductionist Valley

Fuelled by both intrigue and narcissism, humans have a long history of trying to understand themselves.

To comprehend something complex, we humans like to take a subject apart and study what we find inside. By reducing an object to its constituent parts, we attempt to understand the whole.

Early endeavours at reducing biology into discrete components were somewhat fanciful. Hippocrates believed humans were formed of ‘earth, water, air and fire’. Aristotle later argued humans were actually composed of ‘hot, cold, wet and dry’.

As empiricism displaced philosophy, we dismantled life with increasing finesse. In the 16th century, the anatomist Andreas Vesalius saw life as a tapestry of bones, arteries, and tissue. He reduced physiology to “De humani corporis fabrica" (The Fabric of the Human Body). In the 17th century, Robert Hooke observed life beyond the naked eye. Peering down his newfangled microscope, Hooke reduced life into even smaller units. He saw little rooms inside our tissue and called them “cells”.

Empowered by technical breakthroughs in physics and chemistry, 19th and 20th century biologists dove deeper and deeper inside these cells. They were found to contain even smaller biochemical constituents such as proteins, lipids, and nucleic acids. Hereditary units — once imagined as nebulous “gemmules” by Darwin — were reduced to a molecule: DNA.

Life had been reduced to the atomic level.


The history of biological science represents a continued narrative of increasing detail towards a molecular depiction of life. All the way from ‘earth, water, air and fire’ antiquity, to our contemporary molecular vista. This process is a classical example of methodological reductionism.

By sequentially teasing life apart, reductionism has provided a detailed, eloquent and beautiful molecular perspective of biology. It’s humbling to consider the thousands of different molecules we confidently describe with atomic precision. The reductionist view of biology is truly spectacular.

The explanatory potential of reductionism is so seductive, modern biological science has become completely addicted to it. It’s where we live. We’ve travelled deep into reductionism and made it our home.

We live in reductionist valley.

Detail ≠ Understanding

Life in reductionist valley is comfortable. It’s a safe place to empirically explore. But like any comfort zone, it's limiting. I believe our residency in reductionist valley is one of the biggest problems facing modern science — and subsequently cancer research.

The core principle of reductionism is that by understanding discrete components of a system, one can extrapolate to understand the whole. Thus, if we keep cataloguing the molecular components of life, we’ll eventually understand how life works.

Unfortunately, this is only true if the discrete components can be re-compiled to form the larger system we are trying to understand. If we can’t assemble the individual components, reductionism becomes a perpetual high-fidelity cataloguing exercise. Just listing the components of a system — not understanding how the system works.

Limited Lists

Imagine a pile of Lego bricks.

Some pieces are big, some are small. Some are long, some are wide. Some are traditional bricks, some are specialised. A nearby box displays a completed Lego spaceship — but no instructions on how to build it. 

A reductionist can count, measure and document the bricks. They can describe the bricks with absolute accuracy. But ask the reductionist how these bricks combine to build our spaceship, and it becomes less clear. Some bricks look like they might form a wing, maybe others look like the cockpit — but ultimately the reductionist’s list of bricks alone does not contain enough information to build the spaceship. The list is not emergent.

Now imagine our spaceship is a tumour.

A single human cell contains over 20,000 different protein-coding genes, 100,000 protein isoforms, and around 1.0E+10 individual protein molecules. That’s one cell. A small (1 cm3) tumour is composed of approximately 1.0E+9 cells. That means a single tumour contains around 1.0E+19 protein molecules. 

If a small tumour were built from Lego bricks instead of proteins, it would fill 15 Grand Canyons. Most clinical tumours would be the size of a small country. At this scale, the emergent tumour cannot be captured by a reductionist list – no matter how precise.


Cancer biology has spent a long time making lists in reductionist valley. We’ve made base camp our home and devoted much of our exploring to the documenting of more genes, transcripts and proteins.

The challenge of 21st century biology is not to continue accumulating detail in reductionist valley. Our challenge is to use the reductionist detail we’ve gained to explain higher-order biological systems.

We need to escape from reductionist valley.

But with a lifetime experience in digging down, how can we learn to climb?

Fortunately, wading against the reductionist course is more familiar to cancer biologists than they may realise. For example, each time we ‘functionally validate’ an observation, it often involves traversing up the reductionist slope. A differentially regulated mRNA is ‘validated’ by looking at the protein. We align a ‘lower-level’ reductionist observation with a ‘higher-level’ observation to confirm that the reductionist detail empowers an emergent consequence.

Cellular ‘systems biology’ takes this idea a step further. Just as the 20th century scientists used advances in physics to descend into reductionist valley, a modern systems biologist uses mathematical models of ‘omic-level’ reductionist data to explain emergent cellular phenotypes. 

To escape from reductionist valley, we need to continue this ascent.

The Next Ridge

We’re becoming confident climbers when traversing from genes, to proteins, to cells. The next step is steeper and requires new tools again: we need to understand how reductionist events regulate an emergent population of millions of different cells.

The plural of a cancer cell is not a tumour. Mutated cancer cells exist alongside millions of stromal, endothelial and immune cells — each with their own phenotypic potential. Despite this, we often study tumours as an isolated horde of cancer cells. We sequence thousands and thousands of cancer genomes in the hope that a longer reductionist list will somehow explain the emergence of complex heterocellular pathology.

Yet tumours clearly sit atop reductionist valley. They are emergent heterocellular systems. Should we be surprised we can’t understand them from our view at the bottom? 


To understand cancer, and ultimately to treat it more effectively, we need to switch our focus from genocentric reductionism to the emergence of heterocellular phenotypes. Fortunately, a new wave of single-cell phenomic technologies is finally providing the tools to achieve this. From single-cell RNA-seq to single-cell signalling mass-cytometry, we are now starting to measure how molecular reductionist events are regulated across huge populations of cells in a tissue. High-dimensional reductionist detail is being used to understand an emergent system.

It’s early days and we have a lifetime of work ahead. The mathematical challenge of integrating such data will be a field in its own right. But finally the tools exist to at least start our ascent. 

Scientists have always been explorers — and it's the responsibility of explorers to leave their comfort zone.

The time has come to escape from reductionist valley. 


This post originally appeared as a 'Bigger Picture' blog for The Institute of Cancer Research.

Reciprocal Signalling

Cells within a tumour can be broadly classified into two types:

1) Cancer cells (mutated).

2) Stromal cells (not mutated).

It's well established that mutations in cancer cells can drive phenotypic changes within cancer cells. For example, a point mutation in a kinase gene can result in hyperactive signalling — causing a cancer cell to grow too fast. When a cell is affected by its own mutation, it's known as a cell-autonomous event. 

Most genotype-to-phenotype studies in cancer focus on cell-autonomous events. 

But in a real tumour, mutated cancer cells are not alone. They sit alongside healthy, non-mutated stromal cells. In isolation, stromal cells are well behaved. However, in a tumour, stromal cells can be coerced into bad behaviour by cancer cells. They can be bullied by their mutated neighbours.

When a mutation in one cell regulates the behaviour of a neighbouring cell, this is called a non-cell-autonomous event. 

Despite its huge role in cancer, non-cell-autonomous signalling is poorly understood. Most examples are anecdotal to a single pathway or molecule. We actually know very little about how mutations signal beyond cancer cells. It's often presumed to occur, but we have no idea how much, and what the consequences are. 

Why Bully?

Several years ago I started working on how mutations signal beyond cancer cells in pancreatic ductal adenocarcinomas (PDA).

PDA is a nasty disease that contain lots of stromal cells. In fact, some tumours contain even more stromal cells than cancer cells. 

PDA is driven by mutations in a gene called KRAS. I wanted to understand how mutant KRAS signals through PDA cancer cells and stromal cells. If we can understand this bullying process better, we might be able to offer new treatments to patients. 

I started by measuring how KRAS affects cell-autonomous signalling. Using high-throughput multivariate phosphoproteomics, we found that KRAS regulates a very specific section of cancer cell signalling. Mainly MAPK, CDK1/5, and CKII pathways.

Next we looked at non-cell-autonomous KRAS signalling. It's been known for a while that KRAS communicates with stromal cells via a protein called sonic hedgehog (SHH) (no not that one). We found the same thing.

But we also noticed something else: non-cell-autonomous KRAS causes stromal cells to make new, unique growth signals. These signals are distinct to the stromal cells and are not produced by cancer cells on their own. Cancer cell KRAS bullies stromal cells into making new growth signals.

This posited an interesting question: do bullied stromal cells send signals back to the cancer cells? Does KRAS drive reciprocal signalling?


It's easy to hypothesise reciprocal signalling. It's much harder to actually test it. To my knowledge, no one has ever shown mutations can uniquely regulate cancer cells via stromal reciprocal signalling. 

Why? Because it's tricky to see. To observe reciprocal signalling, you need the following:

1) You need to measure cell-autonomous, non-cell-autonomous, and a hypothetical reciprocal signalling state in a single experiment. For this, you need a technique that can measure several phosphoproteomes at once. 


2) You need to know which signal comes from which cell type. That is, you need a technique that provides cell-specific resolution of two heterotypic phosphoproteomes from multiple conditions. 

It's not trivial.

Heterocellular Multivariate Phosphoproteomics

Over the past 4 years I've been developing techniques for multivariate phosphoproteomics and cell-specific isotopic proteome labelling (alongside the CRUK MI Systems Oncology team). I combined these techniques to perform 'heterocellular multivariate phosphoproteomics' (HMP). This technique allowed us to monitor thousands of phosphosites across 10 conditions in two cells at the same time. Just what we needed to test reciprocal signalling.

So what did we find?

When cancer cells non-cell-autonomously communicate with stromal cells, the number of regulated phoshosites in cancer cells doubled relative to cell-autonomous signalling. Cancer cells were getting a big reciprocal signal back from the stromal cells.

Reciprocal KRAS activated major signalling hubs (such as AKT) - regulating transcription, protein abundance, metabolism, cell death, and cancer cell growth. KRAS uniquely regulates cancer cells by bullying its neighbours. 

We put together a little video describing KRAS reciprocal signalling:

Reciprocal signalling is not non-cell-autonomous signalling in reverse. A reciprocal axis starts with an oncogenic mutation, travels through a differentiated stromal cell, and then – using new signals produced by stromal cells – returns to activate pathways the cancer cell cannot activate on its own.

Reciprocal signalling allows mutations to expand their signalling capacity via differentiated heterocellularity. 

This work has just been published over at Cell. If you're into spectra, we've also made all the raw mass-spec data available at PRIDE (PXD003223).

Heterocellular Signalling

The human body contains 40 trillion cells. That's x100 more cells in your body than there are galaxies in the universe. You are a walking cellular multiverse. 

Although your 40 trillion cells all have the same genome, your body actually contains over 200 different types of cell. Each cell type differentially expresses disparate proteins to enable unique emergent phenotypes. Your cells are differentiated

Consider something as simple as your skin. It's not just made of 'skin cells'. Skin is not homocellular. The first 0.5 mm alone contains keratinocytes, melanocytesdendritic cells, merkel cells, and lymphocytes. Tissues are heterocellular

Heterocellularity allows tissues to achieve multiple phenotypes from a common genome. In your skin, keratinocytes form a barrier against the outside world, while your lymphocytes fight off infections. Same genotype, differentiated phenotypes. 

Coordinating heterocellularity requires constant communication between different cell types. This is called 'heterocellular signaling' and it's essential for all metazoan life.

Heterocellular signalling is also frequently dysregulated in disease. Cancer, neurodegenerative disorders, and infectious disease all involve disrupted heterocellular signalling. If we want to combat these conditions, we need to understand how different cells communicate.

Despite the importance of heterocellular signalling, it's experimentally awkward to study.

For example, if you lyse heterocellular mixture of cells and measure the signals using traditional biochemical methods, it's impossible to tell which signal originated from which cell type. This technical limitation makes studying signalling polarity between two cells a laborious process. 

As a result, most people just look at one cell type at time. It's easier. We've learnt a lot that way, but it does mean our understanding of cell-signalling has been experimentally biased towards homocellular models (see A) when biology is actually heterocellular (see B). 

Fortunately, things are starting to change. 

Over the past few years several nascent techniques have emerged that now enable systems biology analysis of heterocellular signalling. These techniques have facilitated pioneering studies of heterocellular signalling and described unique intercellular communication events.

I've curated some of these developments into a new review article for Trends in Biotechnology entitled "Systems Biology Analysis of Heterocellular Signaling".

The review discusses how physical, spatial, and isotopic cell-resolving methodologies are empowering unique studies of heterocellular signalling. If you're interested in studying how cells communicate, take a look. 

Phenotype to Phenotype

Good science is predictive.

When trying to predict how a disease might progress, modern science loves to use genetics. Normal gene X = patient does better, mutated gene X = patient does worse. 

Genotype can predict disease phenotype. But is it the best way?

Cancer is often considered a genetic disease. Mainly by those who can't stop sequencing everything

But while cancer is caused by mutated genes, it's cellular phenotypes that ultimately kill a patient. For example, consider a mutant kinase gene. The mutant gene itself is not pathological (cancer mutations are everywhere). However, the new cellular signaling network produced by the mutant gene can wreak havoc. Deregulated signalling might cause the cell to grow, to move, to avoid the immune system. To spread into the liver and kill a patient. Cellular phenotypes functionally distinguish healthy tissue from tumours. 

But how linear is the route from genotype to phenotype in cancer? We know cancer is genetically heterogenous. What we don’t know is whether this genetic chaos produces equally, less, or more chaotic cellular phenotypes. Genotype-to-phenotype linearity limits the accuracy of extrapolating from genetic data — and we don't understand it. 

So rather than using genetics to predict cancer progression, why not use cellular phenotypes themselves? Why not use cell signaling? Cellular phenotypes are, after all, what defines the disease phenotype. 

A precocious new paper from Fey et al., in Science Signaling does just that. The authors demonstrate JNK pathway flux can be used to predict neuroblastoma survival. Prognosis based not on up-regulation of a biomarker, or the presence of a mutated gene — but how the cell actually processes signals. How signals flow through the cancer cell. They show how a dynamic cell signaling phenotype can predict the larger disease phenotype. Not a gene in sight.  

It's an awesome proof-of-principle. But the question of linearity still remains: which is better to predict disease phenotype? Genetics or cellular phenotypes?

With more eloquent studies of cellular phenotypes — maybe paid for by dropping some dreary sequencing projects — and we might get an answer. 

Packed Lunch

I recently spoke at a 'Packed Lunch' event for The Wellcome Collection.

'Packed Lunches' are midday seminars open to the public  that host the ramblings of a local scientist (often funded by Wellcome). The audience are encouraged to bring their sarnies, an open mind, and whatever questions they fancy. In many ways, 'Packed Lunches' are like traditional academic lunchtime seminars, reformatted for the 'public communication of science'. 

However, unlike typical lecture-style talks, 'Packed Lunches' are public interviews. There are no slides. No laser pointers. 

Frankly, I was surprised Wellcome invited me. They normally interview lab-heads and professors. They'd clearly run out and needed to draft from the B-team. I almost didn't do it for fear of looking unqualified. I was equally surprised that anyone (beyond my parents) came.

In the hypercritical environment of modern research, it can be easy to forget we scientists do something others care about. One of the joys of public engagement is being reminded  that even if all your experiments failed and your paper was just rejected  there are people out there who care you're even trying.

However, being interviewed (and recorded) in-front of a live non-scientific studio audience was very alien to me. A public interview  with no narrative control  is an unsettling prospect. I was terrified.

Fortunately, the interviewer, Tom Anthony from Wellcome, kept the questions simple and my jargon leashed.

Yesterday, Wellcome released the event as a podcast. Unfortunately they don't put podcasts on iTunes (I've no idea why, hopefully this will change). So if you'd like to hear me ramble, a recording of the event can be found below:

Creative Types

As a part of my Wellcome Trust postdoc I've spent the last month working as a researcher for the BBC Science Division. Given this is an unconventional way for a scientist to spend their time, I thought I'd share some insights from inside the BBC.

On Location

The BBC Science Division is based in New Broadcasting House, London (AKA W1A). To my lab-weary eyes, the BBC headquarters makes an exciting first impression. It's a colourful hive of wide-eyed creativity and celebratory heritage. Offices are named after classic BBC locations (I worked in Nelson Mandela House). Our team meeting room was inspired by The Queen Vic. Even the elevators proudly broadcast BBC radio. (Note: The 1Xtra lift is genuinely stressful. There's a reason why grime isn't normally played in claustrophobic boxes transporting the middle-class.)

It's an exciting environment. 

But beyond the glossy media facade and my quest to find all the Dr. Who meeting rooms  there was work to do. And I was lucky enough to work with the best: BBC Horizon

I've always been a fan of Horizon. With a 50-year heritage and an international audience, it's arguably performed more science communication than any other production. I have a huge amount of respect for anyone who popularises science well. Horizon are some of the best.

It's a diverse show and my time at Horizon reflected that. One day I would be researching private moon exploration the next, local science-community relations. Some days you're hunched over a laptop in the office  the next, you're traversing London with a film crew. 

Fortunately I got to see the the whole process. From initial idea conception, research and treatment writing  all the way through to filming, editing and sound mixing. I even managed to get my dad on telly.

Obviously, rambling with film crews and ruminating on editing sofas is unlike anything I normally do in the lab. However, I went to Horizon knowing that (unless they give me my own TV show) I'll still want to be a scientist afterwards. So why the sojourn?

What did I learn from the BBC?  


Well firstly, I learnt that communicating science on TV requires a respect for the interests of millions of people. Most scientists don't think on this scale. We care about our peers, our field, maybe our 100 twitter followers. But that's it. We rarely think: 'what would 2 million people think about this?' Making a show that the entire nation can watch re-calibrates priorities. 

For the most part, this general-public appreciation was insightful. It was a healthy, empathetic exercise.

However, for me, public-consideration had one disturbing outcome: 'audience-value' is often prioritised over 'scientific-value'. That is, the televisual value of scientific content is judged by whether the public are already interested in it. The presumption being that no one will watch a science programme on a topic they don't already care about. 

This 'audience-value' > 'scientific-value' paradigm was the hardest thing for me to stomach at Horizon. Pitching idea 'X' would often raise the following audience imitation: 'Yes, but it's 9pm on a Tuesday. I've been at work all day. I'm tired. Do I really want to watch a show about 'X'?'. I understand why TV-professionals think like this. There are 500 channels of competition, and ratings matter. 

But it's this thought process that produces 'science' shows about cats

I think the challenge of science communication is to make people care about something they don't know they care about. Or as Oliver Burkeman says of writing generally: "The idea is to help them [the audience] discern something you know they'd be able to see, if only they were looking in the right place".

Not to educate people on what they already care about.

Scientific communication should better than that. It's not about topping up knowledge. It's about raising consciousness

The target isn't public interest. The target is public ignorance. And it's the job of science communication to quench ignorance using the narrative process – not because a pre-existing narrative interest already exists.

Carl Sagan didn't care about how tired you were on a Tuesday night. He aspired to make something so compelling, and so beautiful, that 35 years later we'd still be talking about it. I wasn't looking for how beautiful the universe could be until he showed me. That should be the goal. 

Not every Horizon episode falls foul of this. For example, recent episodes such as 'Secrets of the Solar System' and 'Dancing in the Dark - The End of Physics?' discuss topics with very little inherent public interest. In particular, 'Aftershock: The Hunt for Gravitational Waves' even showed the scientific method in a way I've been craving for years

So it's not all bad. I just want Horizon to be better than that. Not crippled by ratings-pragmatism.

But I'm just one of the two-million people they're writing for and I don't like cats. No single opinion can represent the audience. 


Surrounded by all these media-folk, I also achieved a new appreciation for what creativity means in science. 

I've always thought of media people as 'creative-types'. Compared to us drab scientists they're contemporary, fashionable, and flamboyant. They have neo-ironic clothes, calculated haircuts and Instagram addictions. They design, produce, and create. They're bohemic right-brainers. 

Scientists conversely – reside at the other end of the spectrum. We're perceived as meticulous, comprehensive, insipid-types. Our beards are sincere. Our glasses prescribed. Straight-up left-brainers. We can't possibly be creative. 

This dichotomy is utter bullshit.

For a start, the BBC is full of nerds. Producing world-leading television is not a job for materialistic hipsters. This is straight-A territory. These media-guys are incredibly smart. 

Secondly, far from lacking creativity  in the context of factual TV scientists are the creativity. Scientists are the content. Scientists produce the stories. Without scientists there is nothing to film. 

Whether they realise it or not (I certainly didn't), scientists spend their entire lives creating content and media-types like those at Horizon need it

Re-packaging scientific content into good factual TV-shows is really hard. I've got a new-found respect for how hard it is. Producing 15 hour-long factual documentaries a year to the quality and diversity Horizon achieves is seriously impressive. 

But just because they author fancy graphics, orchestral scores and celebrity voice-overs  the media are not the 'creatives' in scientific communication. They are (very good) reporters. 

So if I've learnt anything from the BBC Science Division it's this: scientists are the true creatives

The Imitation Game

UK higher-education research institutions were recently critiqued by the Research Excellence Framework (REF). The REF aggregates individual researcher performance by host institution to produce an overall score of institutional prowess. This allows qualitative comparison between research institutions as a product staff output.

Atop this process is £2 billion in research capital. The aim of the REF is to “distribute these funds selectively on the basis of quality”. That is: do well in the REF and you get more of the £2B. This meritocratic financial outcome means REF results are hugely important for research institutions. 

Recently, the pressure to obtain a high REF score has led to accusations of systematic ‘gaming’. That is, finding loopholes in the REF process to obtain better scores. One well-publicised ‘gaming’ technique is ‘selective submission’.

The technique is very simple: as institutional REF scores are an aggregate of all submitted individuals, some institutions knowingly submit discerning individuals and fail to submit those who might drag the score down. This falsely biases the intuitional score up. For example, Cardiff University only submitted 62% of REF-eligible staff. They selectively picked their best staff to get the best institutional score. Fortunately, as we have a record of how many REF-eligible individuals each institution should be submitting, it’s possible to quantify the ‘selective submission’ gaming process and correct for it. This recently happened with a revised REF scoring. (Cardiff dropped from 7th to 50th.) 

Correcting for REF gaming is crucial to obtain an accurate quantification of institutional prowess. It ensures funds are fairly distributed and allows future employees to empirically select the best places to work.

There is another form of REF gaming that receives less attention that I think we should also be correcting for....

I call it ‘output imitation’.

Here's the premise: REF submitted research does not have to be performed at the institution which it represents. For example, all 3 REF papers I submitted for my current institution (ICR) were performed in my Ph.D. lab (Cambridge). That means the ICR gets judged for work done in Cambridge

If the REF aims to measure individuals – that is fine. But given the REF purports to measure institutional research quality, this opaque allowance is extremely deceiving. 

Output imitation ensures REF scores do not represent the quality of work done at a host institution. Instead, REF scores represent a combination of work that was done at a host institution and work that was imported from elsewhere.

I believe the REF should resolve between these two research types.


Because it's much easier to import research papers from elsewhere  than to provide the resources, time, money and stability so that equivalent research can be performed in-house. 

It's widely accepted that junior researchers are selectively recruited because they can transfer sexy papers from other labs. Meritocratic hiring is great – but only if the host institution is sincere in what they offer junior researchers.

The prevailing (anecdotal) zeitgeist implies successful junior researchers are recruited to bump the institutional REF score  and then left under-supported when it comes to producing new research. Why would an institution spend time and money supporting existing talent when it can easily import a new wave of successful juniors for the next REF? 

Output imitation allows research institutions to overcome poor internal investment by importing external merit. It encourages insincere short-term hiring and discourages long-term internal investment.

I propose a single checkbox next to every future REF paper submission: 'Was this research performed at the host institution?'

Yes or No.

If the REF could institutionally resolved between those supporting internal researchers and those recruiting external prowess, host institutions will have an incentive to actually support the talent they recruit. 


Disclaimer: The ICR came 1st in the REF. I've got no personal reason to complain about the current system. I just think it's deceiving. 

Creative Crash

Academic scientists perform research to expand the boundaries of human knowledge. New knowledge may have application – it may not  but ultimately, academics pursue questions because they are interesting

Academics are 'blue sky' researchers. They value intrigue over profitable application. 

In contrast, industrial scientists (working for companies) use the scientific method to make money. They perform research as a means to capitalism. They pursue questions because the answers are profitable

Industrialists are 'profit maximisation' researchers. They value profitable application over intrigue. 

Each axiom also has its trade-offs. Academics accept reduced professional and financial stability in exchange for creative freedom. Conversely, industrialists accept reduced intellectual freedom in exchange for professional and financial stability. 

Both academic and industrial scientists are well trained. Both are smart. Both do cutting-edge research. One is not 'better' than the other. Academic and industrial scientists simply use the scientific method for different goals based on whether they value intrigue or profit respectively.  

This professional dichotomy allows scientists to choose between the creative freedom of an academic career or the profitable stability of an industrial career.

At least that's how it's been in the past – because this paradigm appears to be changing. 

Recently I've observed increasing claims that a 'creative crash' in academic science is looming.

Writing for Ars TechnicaBen McNeil explains:

"The hundreds of billions of dollars of government funding that supports the world's academic research ecosystem is distributed based almost exclusively on the opinions of senior experts (or ‘peers’). These experts review proposals and seek to find ideas impervious to criticism. Unfortunately, a research idea that is immune to criticism during peer review will, by its very nature, be cautious and take minimal risks.

Rather than have peers assess the innovative potential of an idea, preliminary data and publication records are now the dominant parts of the evaluation. Funding is so tight and proposals are so heavily critiqued that any one reviewer can kill a grant proposal based on arbitrary metrics of quality—or even if they suspect the idea just won’t work.

Yet relying only on peer-review misses something about the nature of scientific innovation: some of the biggest discoveries are deemed crazy or impossible by experts at the time."

Risk-adverse critique by senior peers has the potential to produce a conservative (arguably right-wing) anti-novelty bias in scientific funding. Given that new ideas are the core medium of academic research, this anti-risk taking bias is extremely dangerous.

New ideas  by definition  always journey into the unknown. The entire purpose of non-profit academic research is to push the boundaries of our knowledge. But you can't push boundaries if you're scared things won't work and want to play it safe. Famously: If we knew what we were doing it wouldn’t be called research, would it?

New ideas are high-risk. That’s exactly why professional companies can’t justify doing them. And that’s exactly why non-profit academics should be doing them.

What is the point of academia if only low-risk, highly-proven projects are the only ones who get funded? If academics can’t pursue creative, novel, blue-sky research  yet still experience volatile employment conditions  the traditional academic (freedom + instability) vs. industrial (stability + no freedom) argument unbalances. Academics might as well work in industry. 

Unsurprisingly it's already happening.

I've been coming across more and more and more and more examples of well-trained, successful scientists leaving academia because they couldn't get funding for creative research. These people are not lacklustre bums. These are people who have been educated into their 30's by the very funding sources that are now crippling them. And because they lack the piles of preliminary data and safe track-record that grant reviewers require, the conservative funding bias can hit junior researchers hardest. 

Ben McNeil recounts an increasingly familiar perspective:

"In the early 1970s, Roger Kornberg, a 27-year-old Stanford PhD, was working at the Laboratory of Molecular Biology in Cambridge, England. With a modest post-doctoral salary, Kornberg was given freedom to explore untried and risky areas of research. This would ultimately allow him to make a revolutionary discovery about how DNA is copied in cells.

Kornberg would win the Nobel Prize in chemistry in 2006 for that work. Yet he told The Washington Post that he’s convinced his groundbreaking Nobel Prize winning idea would never have been funded today.

"If the work that you propose to do isn't virtually certain of success, then it won't be funded,” he said. “Of course, the kind of work that we would most like to see take place, which is groundbreaking and innovative, lies at the other extreme.""

And Kornberg isn't the only Nobel Prize winner who's worried about research funding. Sydney Brenner shares a similar concern

Fortunately, all is not lost for those wacky ideas just yet. A small rebellion has begun....

The Wellcome Trust recently announced their new 'Seeds' funding. A Seed grant is "a new kind of funding to support the generation of new ideas. It aims to address a gap identified by our community: small awards to investigate riskier concepts." Seeds are explicitly for un-proven creative ideas. Wellcome can do this because they're rather moneyed and don't have an explicit applied science agenda – but there is little reason why other grant sources can't attempt the same.

Hopefully more funding bodies will remember their academic objective is to push the boundaries of human knowledge. Not to tip-toe around the boundaries  scared incase something doesn't work.