New machine learning models to identify driver mutations for clonal hematopoiesis

Clonal hematopoiesis is a process in which some blood stem cells acquire mutations that allow them to grow and multiply faster than others. This favours a predominance of cells carrying these mutations in the blood. Clonal hematopoiesis is a phenomenon that becomes more common as we age and it can increase the risk of developing serious diseases such as hematologic cancers and cardiovascular diseases. 

Identifying the mutations that cause this condition is crucial to better understand how they affect our health and to develop ways to monitor people at higher risk of developing these diseases.

In 2022, the Biomedical Genomics laboratory at IRB Barcelona identified 64 genes linked to clonal hematopoiesis. Now, the team, led by ICREA researcher Dr. Núria López-Bigas and Dr. Abel González-Pérez has published in the journal Cancer Discovery a series of models based on machine learning that identify the specific mutations—in 12 of the most commonly altered genes—that drive clonal hematopoiesis. 

These models are more precise than  expert-curated rules currently used to identify these mutations as they can uncover patterns in the mutations that were not evident previously.

“On the one hand, these models will allow us to better understand the mechanisms underlying clonal hematopoiesis, but also in the future, they may open new opportunities to monitor and treat people at risk of developing diseases associated with this condition,” says Dr. González-Pérez.

Results in large patient databases

To validate the computational models, the team used a database of almost half a million individuals from the UK Biobank. Thus, they were able to identify individuals with clonal hematopoiesis and confirm known associations between this process and age, or other risk factors, such as having previously received chemotherapy treatment for cancer, or smoking. They also confirmed associations with the risk of suffering from other diseases, such as blood cancer or cardiovascular disease.

“The results show that these models are more specific and sensitive than expert-curated rules based on prior knowledge that are traditionally used, thereby highlighting their potential to identify new mutations that can cause clonal hematopoiesis,” explains Dr. Joan Enric Ramis-Zaldivar, one of the first authors of the work, together with Dr. Santiago Demajo and Dr. Ferran Muiños.

A Parallel Development to that of Cancer

Clonal hematopoiesis, like the development of tumours, consists of accelerated cell growth. In the case of cancer, cancer cells proliferate uncontrollably and acquire new capabilities such as invasion of adjacent tissues and migration. In clonal hematopoiesis, certain blood stem cells gain a growth advantage and expand. 

In both cases, this expansion is caused by one or more mutations, which confer a certain advantage to the cells, namely the capacity to grow and reproduce faster than their neighbours. Clonal hematopoiesis is considered to occur when 2% of all blood cells in an individual come from a single original cell, meaning they are clones.

The researchers have adapted the tool BoostDM, developed by the same laboratory to study cancer driver mutations, to address clonal hematopoiesis, applying advanced models and algorithms to identify mutations that promote this process.

This study is part of two research projects that the group is conducting in collaboration with the Spanish National Center for Cardiovascular Research (CNIC). These projects aim to study the mechanisms that link clonal hematopoiesis to cardiovascular diseases such as atherosclerosis and myocardial infarction. 

The research has been possible thanks to the support of multiple funding agencies, including the European Research Council, the Spanish Association Against Cancer (AECC), the Spanish Ministry of Science, Innovation and Universities, the PROMINENT project of Cancer Grand Challenges, and the CGI-Clinics project within the EU Horizon Europe Programme, among others.

Future work by the laboratory will focus on the development of new models that cover the mutations of more genes linked to clonal hematopoiesis through the analysis of new databases.

Related article:
Identification of Clonal Hematopoiesis Driver Mutations through In Silico Saturation Mutagenesis 
Santiago Demajoº, Joan Enric Ramis-Zaldivarº, Ferran Muiñosº, Miguel L Grau, Maria Andrianova, Núria López-Bigas* & Abel González-Pérez*
Cancer Discovery (2024) DOI: 10.1158/2159-8290.CD-23-1416