A Meta-Analysis of Cancer Metabolism Helps ID Drug Targets and Predict Disease Severity
Data Reuse Digest: 11-20-23
Introduction
Welcome back to the Data Reuse Digest - and the latest edition of our series on computational research for type II diabetes. We are exploring how different kinds of computational studies can drive clinical progress. By sharing success stories in one disease area, we aim to inspire the implementation of these successful approaches for other diseases also.
If you have friends or colleagues that you think would benefit from this newsletter, you can share it with them by clicking this button:
And if you are reading this now, but you’re not yet a subscriber, you can subscribe by clicking this button:
A Meta-Analysis of Cancer Metabolism Helps ID Drug Targets and Predict Disease Severity
Cancer cells are built different. They can evade the immune system, grow and multiply uncontrollably, and spread throughout the body. To understand the root of these deviant behaviors requires picking apart the unique properties of cancer cells at the molecular level. It is a task that many scientists devote their careers to. In this edition we will focus on one effort to sum up what we know about cancer cell metabolism. The implied goal of this and other studies is to identify vulnerabilities in the cancer cell’s machinery that can be exploited for treatment.
What is metabolism? In everyday conversation we like to talk about ‘fast metabolism’, ‘slow metabolism’, or ‘hacking our metabolism’. All these phrases revolve around the idea that metabolism has something to do the way we eat and process food. Indeed, metabolism as a scientific concept does involve the processing of food – but within the scientific community the metabolism is more commonly considered from a cell biology perspective. From this viewpoint, the role of your digestive system is to break down and shuttle nutrients into the blood stream so that they can be transported to all the body’s many different cells – including cancer cells. Inside each individual cell, a complex series of chemical reactions takes place to break down these nutrients into usable components and generate energy so the cell can carry out its business.
It has long been known that the metabolic machinery in cancer cells works differently from regular cells. Way back in the 1920s, German physiologist and Nobel laureate Otto Warburg observed that cancer cells increase their uptake of glucose (sugar), converting relatively more glucose to the metabolite lactate (a metabolite is simply the product of a metabolic reaction) than normal cells. This phenomenon is known as the Warburg Effect, and scientists are still studying it a century onA.
These metabolic reactions – the uptake of glucose, the conversion of glucose to lactate – are very well studied. But cellular metabolism involves thousands of chemical reactions and metabolite productsB. What about the rest? How could they contribute to the activities of cancer cells? Modern scientists have at their disposal lots of fancy machinery capable of measuring hundreds or even thousands of cellular metabolites all at onceC. In prior studies, scientists have used these machines to compare the levels of cellular metabolites in cancer patients and healthy individuals. The data from these studies is present in public repositories.
There’s now enough of this data available for someone to sift through it and make sense of it all: what prior findings are we confident about, which are we less certain to be true, and which are deserving of further follow-up experiments. These are the goals of a ‘meta-analysis’ – to describe what we know and lay out a potential roadmap for what we ought to study next.
In this edition’s featured study, Reznik et al. compiled data from 11 different studies, covering 7 different cancer types. As a whole, these 11 studies gathered 928 tissue samples from both healthy individuals and patients with different kinds of cancersD .
Featured Study: A Landscape of Metabolic Variation across Tumor Types (Reznik et al., 2018, Cell Systems)
Instead of diving into the specific details of the meta-analysis – how the researchers combined the datasets and analyzed the data – or listing out all the specific metabolites associated with cancer, I will instead summarize the key questions that researchers asked about the data to give a sense of what a meta-analysis can uncover (for more details on the meta-analysis methods or specific metabolites, you should check out the paper and play around with the very nice web application that the researchers designed to share their data)E.
First, the researchers looked at all the different kinds of cancers individually. For example, they asked the question: to what extent is kidney tumor tissue different from healthy tissue? They found that for certain cancers (including kidney cancer and breast cancer), over 60% of metabolites showed significantly different levels compared to healthy tissue. Other cancers were less metabolically disturbed. Why certain cancers seem to have more abnormal metabolism than others is an open question, but one that might shed light on the different strategies that cancers use to grow and their vulnerabilities to certain treatments.
In addition to comparing the different cancer types, the researchers also looked for features that united them. They identified 27 different metabolites that were consistently differentially abundant (either found at much higher or much lower levels in cancer cells than normal cells) in at least 6 of the different cancer studies.
These results are certainty interesting for what they tell us about cancer biology, but the researchers wanted to go further and identify potential therapeutic opportunities. To do so, they identified enzymes that could be producing or consuming the metabolites that the researchers found associated with different cancer types (using a computational tool called ‘paxtoolsr’). Of these enzymes, the researchers found 57 that are targetable by drugs either already approved by the FDA or currently under investigationF. This means that these drugs may be repurposed to target the cancers investigated in this study. The idea is that if you can mess with their metabolic machinery, you may be able to curb the growth of cancer cells and make them more vulnerable to other treatments. In addition to repurposing drugs, another objective of translational research is to predict disease severity given a patient’s biological data. The study authors did this as well, showing that a variety of metabolites are associated the grade (severity) of a tumor. Some of these metabolites are common predictors of severity across multiple cancer types.
In addition to these questions, the researchers asked even more, which you can explore by reading the paper. I hope you see the potential for meta-analysis studies like this one to cement our existing biological knowledge, uncover new biological details, and point the way towards new therapeutic possibilities.
Research Trends
The point of this section is to provide big-picture context: how are the featured studies shared in this edition representative of broader trends in computational research? These trends will be sometimes cite information from past editions, additional research articles, and mainstream news stories.
(A) One very recent, high-profile paper (Bartman et al., Nature, 2023) references the Warburg effect. The authors show that the rate of glycolysis is increased in tumors compared to healthy tissue, as the Warburg Effect suggests. However, the paper adds nuance. While past researchers have assumed that cancer cells are ‘hypermetabolic’ – they consume glucose and produce energy at a more rapid pace – the researchers show that cancer cells in fact produce energy (in the form of the energy-carrying molecule ATP) at a slower rate. The cancer cells compensate by synthesizing proteins more slowly. The researchers suggest that cancer cells grow rapidly by ‘shedding’ certain functions that are not necessary for their growth. In the discussion section of their paper, the authors suggest that efforts to further manipulate ATP production may help slow cancer growth.
(B) The number of metabolic reactions that are taking place in human cells is still not well defined, but researchers have previously estimated the number of metabolic enzymes operating throughout the human body – for example, from one study: “The 2,742 enzyme genes in HumanCyc correspond to 9.5% of the human genome, and can be subdivided into 1,653 metabolic enzymes, plus 1,089 nonmetabolic enzymes (including enzymes whose substrates are macromolecules, such as protein kinases and DNA polymerases)” (Romero et al., Genome Biology, 2005).
Researchers are still actively cataloguing the full range of human metabolic functions. And there are other open questions – like to what extent different metabolic functions are operating in various cell types, including cancer cells as in the study from (A), or how metabolism differs in diseased vs. healthy tissue (say in people with diabetes vs. people who don’t have the disease)
(C) The two most common techniques for metabolic measurement are mass spectrometry and NMR (Nuclear Magnetic Resonance) spectroscopy. There are a number of variants of each technique.
Here is a great review article that discusses techniques for metabolomics data collection, processing, and analysis (Jang et al,. Cell, 2018). And here are some basic intro videos about the fundamentals of ‘Mass Spec’ and NMR if you want to learn more about them.
(D) It is easy to get lost in the numbers, but worth remembering that each of these tissue samples came from individual people – cancer patients, likely making frequent trips to the hospital for treatment, and taking even more time out of their lives to participate in a clinical trial (some of them may be participating in multiple at once). They are, by donating their biological ‘data’ to the research community, making better treatment possible for future cancer patients. Without them there would be no cancer research.
For an excellent book on the history of cancer research and the patients who have made it possible, check out Siddhartha Mukherjee’s Pulitzer-Prize Winning ‘The Emperor of All Maladies’. It is a bit of a lengthy read but worth the time (in grabbing a link to the book, I came across the Ken Burns documentary of the same name – a good supplement / alternative to Mukherjee’s text). Mukherjee’s other work is very well done – for example, his latest on the history of cell biology and cellular therapies, ‘The Song of the Cell’.
(E) These research web applications are a valuable resource in the age of -omics research – studies that measure the expression of thousands of genes, proteins, or metabolites and compare them across patients. There is simply too much data to effectively discuss inside the text of an academic publication – and trying to summarize the full range of results often comes with a loss of clarity.
Research web apps allow readers of the paper, each with their own biological background and interests, to peruse the data selectively, seeking out their own target molecules and biological systems of interest in the data. The creation of these web apps seems to be a growing trend, at least from what I can see in my reading of the literature – here’s another example of a paper from this year (Julkunen et al., 2023) with its own web app. This one, named the ‘Nightingale Atlas’ after the woman who founded modern nursing, presents an analysis of blood biomarker data from over 100,000 individuals in the UK biobank (another database whose data is very often re-analyzed in research studies)
(F) This demonstrates the value, yet again, of large public databases of biological data – and software tools developed to retrieve their data. In this case, the researchers used the R package ‘paxtoolsR’, mentioned in the text, to query DrugBank. DrugBank was started in 2006 by Canadian researcher Dr. David Wishart’s lab at the University of Alberta. It has grown in the succeeding years to contain detailed information on more than 500,000 drug compounds. There is an incredible amount of detail – as you can see in the entry for diabetic drug metformin. This is knowledge compiled by numerous researchers over many decades of focused work.
But as you can imagine it is hard to make sense of all this data in a meaningful way. Fortunately, the designers of databases like DrugBank design APIs – Application Programming Interfaces – so that database users can extract certain information from the database in an efficient way for many drugs at once. The developers of paxtoolsR took advantage of the accessible database architecture to develop a tool that makes it even easier for fellow researchers to access the data – just install the paxtoolsR package in RStudio and run its defined functions as the tutorial instructs.
This task does require, however, a working knowledge of R and RStudio – the desktop integrated development environment (IDE) that makes it relatively easy to write R code, view outputs, and save scripts of code to run again later. An ability to use these tools is increasingly a requirement for scientists in today’s age, even for researchers who do bench work and don’t work at a computer full time.