Turning a Phylogenetics Algorithm into a Crime Scene Narrative

 Behind the scenes of  "The Tell-Tale Copy Number" 

The Challenge

When I saw the call for the Université de Sherbrooke’s science popularization contest, I immediately thought of one of my supervisor's latest paper—a theoretical algorithm for comparing tumor cell genomes. But how do you turn a 12-page article into something readable, in just 600 words, with no equations, and one image? That was the challenge.

Here you will find the transformation from my initial brainstorm to the final piece, and the design choices I made along the way. Why I ditched metaphors about family trees, embraced true crime podcasts, and chose clarity over technical detail.

Defining the message

Keeping the algorithm, trimming the math 

The silent killer and the tell-tale copy number

Cancer, also known as “The silent killer”, is a disease that starts when some of our cells begin 

to grow chaotically. Usually, our body decides when and how many cells must be produced. 

However, cancer cells want to reproduce themselves indefinitely and our bodies are unable to 

stop them. This ruthless production of cells agglomerates into a bulk of cells which we call a 

tumor. To understand how a single malignant cell can evolve into a tumor, we use 

phylogenetics, a branch of evolutionary biology that studies the relationships between 

individuals over time. Finding these relationships is like a crime mystery; where we take a 

picture of the crime scene – the DNA of the tumor cells – and use it to infer the events that 

happened in the past. To investigate this crime scene Dr. Lafond et.al. proposed

a new 

model that recreates this mystery by using the numerical footprints left by the genes of the 

involved cells. Using phylogenies to study the mechanisms behind cancer, we can not only 

understand, but also predict and perhaps control the progression of cancer and prevent it from 

spreading into other parts of the body.

The clue: The tell-tale copy number

Let’s start with the crime evidence. Every cell in our body carries a certain number of copies 

from each gene. This number of copies per gene in a cell is known as Copy Number Profile 

(CNP). Since cancer cells reproduce indefinitely, they spawn many offspring cells, each with 

their own personality traits. The CNP is one of these traits: the children of a cell might have 

more (amplification) or less (deletion) copies of a gene. To measure the similarity between two 

cells, we can count the number of mutations required to explain the differences in their CNPs. 

Comparing CNPs gives an evolutionary distance between cells, which then can be used to infer 

all the evolutionary relationships within the tumor

Our detective: The cnp2cnp algorithm

Traditional cell comparison models assume that each gene undergoes mutations independently. 

However, recent evidence shows that amplifications and deletions can affect large blocks of 

genes. In this work, a new model - a mathematical representation of the biological problem –

which considers this novel evidence was proposed. The key question in our crime investigation

is: Given two CNPs, namely a source and a target, what was the sequence of events that could 

have occurred to transform the source into the target? Moreover, some amplification or deletion 

events are more probable than others, so is it possible to find the series of events that represents 

the most likely scenario? In this article, it was proved that this problem requires calculations 

that are too difficult even for the most sophisticated computers, which is what computer 

scientists call an NP-hard problem. However, the authors have developed a new algorithm 

named cnp2cnp which is shown to find evolutionary scenarios that are very close to the optimal.

The model was compared against previously used implementations such as MEDICC, which 

considers a restricted form of a gene amplification, and the Euclidean distance, which only 

considers the absolute differences between gene copy numbers. Since the cnp2cnp model is 

closer to the biological reality, it was shown to reconstruct tumor phylogenies with equal or 

superior accuracy.

The end?

Throughout its development, the silent killer leaves a trace of useful evidence from which we 

can retrace and understand its behavior. CNP for example, is a tell-tale piece of evidence that 

can help us infer the evolutionary history of a tumor. In particular, the cnp2cnp algorithm 

makes it possible to incorporate new helpful biological evidence to recreate the events more 

akin to the reality. Adapting current models to more complex phenomena places us one step 

ahead of this silent killer, in hope that one day we will be able to finally stop it