Peter Van Loo: Projects

The advent and exponential cost decrease of massively parallel sequencing technologies over the past years has enabled sequencing entire cancer genomes. This resulted in unique opportunities for cancer research. Large-scale consortia (TCGA, The Cancer Genome Atlas, and ICGC, the International Cancer Genome Consortium) have produced whole genome sequences of thousands of cancer genomes, and are making their data available to the community. These efforts are now further scaling up, as exemplified by Genomics England's 100,000 Genomes Project.

I argue that we have so far only skimmed the surface of what can be learned from this unprecedented wealth of data. There is therefore a clear need for in-depth large-scale pan-cancer analyses. Our group focuses on integrative analyses of large-scale public 'omics data, leveraging the wealth of cancer genomics data into large-scale pan-cancer analyses to understand carcinogenesis and cancer evolution.

Characterising the landscape of tumour suppressors

Many cancer genes are somatically altered in only a very low proportion of tumours, providing a clear rationale for large-scale pan-cancer analyses of driver mutations. We are applying approaches centred on copy number analysis to characterise the landscape of tumour suppressors.

Figure 1

Figure 1: Copy number analysis of cancer genomes. The ASCAT method infers tumour purity and ploidy, allowing detailed inference of copy number profiles, LOH and homozygous deletions. (Click to view larger image)

Many tumour suppressors are targeted by homozygous deletions, removing both parental copies. Because any homozygous deletion that includes a gene that confers a survival advantage is eliminated by negative selection, homozygous deletions are rare and often focal.

Admixture of normal cells in tumour samples has historically hindered the reliable identification of homozygous deletions. We previously developed ASCAT (Allele-Specific Copy number Analysis of Tumours), a method to derive copy number profiles of tumour cells, accounting for normal cell admixture and tumour aneuploidy (Figure 1).

Methods such as ASCAT can now effectively deconvolute copy number profiles of tumour cells from those of admixed normal cells and reliably identify homozygous deletions in tumour samples.

We are applying ASCAT to thousands of samples across cancer types, to screen for tumour suppressors through recurrent homozygous deletions. For a subset of these cases, point mutation data, gene expression data and/or DNA methylation data will also be available, which we will correlate with detected homozygous (and hemizygous) deletions, allowing us to more clearly delineate target genes within regions of homozygous deletions. Through this screen, we aim to characterise the landscape of tumour suppressors and particularly identify rare tumour suppressors.

Tumour suppressors can also be inactivated by a combination of a deleterious germline variant, combined with loss-of-heterozygosity (LOH) of the other allele. We are performing a large-scale pan-cancer screen for this combination of events, as a complementary method to chart the landscape of tumour suppressors.

Molecular archaeology of cancer: inferring timelines of cancer development and evolution

The cancer genome contains within it an archaeological record of its past, and we previously pioneered methods to disentangle a cancer's life history from sequencing data (Fig. 2). We anticipate that a large-scale pan-cancer approach to obtain detailed evolutionary histories of tumours would give profound insights into carcinogenesis and cancer evolution.

Figure 2

Figure 2. Molecular archaeology of cancer: an example. From the picture on the right, one can infer that the purple mutations happened first, then the blue chromosome duplicated, and then the yellow mutations occurred. In addition, from the relative numbers of yellow and purple mutations, one can infer when in the tumours lifetime the blue chromosome duplicated. (Click to view larger image)

We can construct life histories of thousands of tumours from their genome sequences, using both driver and passenger mutations. By obtaining detailed timelines of many cancers' evolutionary histories that include driver mutations, copy number changes, rearrangements and mutational processes, we will identify the initiating events of cancer development, and the events that are selected for later in a cancer's lifetime, including those that drive late clonal expansions and that may play a role in tumour malignancy. In addition, these analyses will allow blueprints of the subclonal architecture across cancer types in unprecedented detail and on an unprecedented number of cases, allowing a glimpse into a tumour's future.

Complementary to this, we are performing smaller-scale collaborative studies of tumour bulk sequencing, in combination with single-cell and multi-sample sequencing of primary tumours, metastases and circulating and disseminated tumour cells, aiming to gain insight into tumour evolution and metastasis.

Deconvoluting expression in tumour and normal cells

We aim to understand how changes to genome lead to transcriptomic changes to finally cause cancer. Deep understanding of the cancer transcriptome is confounded by expression signals originating from admixed normal cells. Gene expression analysis by massively parallel sequencing (RNAseq) allows allele-specific expression measurements. It can be shown that, given the fraction of tumour cells, the allele-specific copy number profiles of the tumour cells, and under a few reasonable hypotheses, the expression in tumour cells can be separated from that in normal cells (Fig. 3). We aim to develop such bioinformatics approaches to deconvolute the tumour cell transcriptomes from transcriptomes of admixed normal cells.

Figure 3

Figure 3: Principle of a method to deconvolute the tumour cell transcriptomes from transcriptomes of admixed normal cells, using copy number data and (allele-specific) expression from RNAseq. (Click to view larger image)

We will apply these methods to large pan-cancer RNAseq datasets, allowing a transcriptome-wide view of cancer across cancer types. We expect these tumour-cell-specific expression profiles will result in a better taxonomy of cancer than mixed-cell-population expression profiles. Finally, expression profiles of admixed normal cells will allow insight into the cellular composition and transcriptional state of the tumour stroma.

In the longer term, we aim to develop integrative genomics-transcriptomics approaches that study the influence of point mutations, copy number changes and structural variants on transcription at the gene or transcript level and at the transcriptome level, and to apply these approaches in a large-scale pan-cancer setting, to understand the basic principles of cancer development and cancer evolution within and across tumour types.

Peter Van Loo
+44 (0)20 379 61719

  • Qualifications and history
  • 2008 PhD in Medical Sciences, University of Leuven, Belgium
  • 2008 Postdoctoral Fellow, Institute for Cancer Research, University of Oslo, Norway
  • 2009 Postdoctoral Fellow, VIB and University of Leuven, Belgium
  • 2010 Postdoctoral Fellow, Cancer Genome Project, Wellcome Trust Sanger Institute, Cambridge, UK
  • 2014 Established lab at the London Research Institute, Cancer Research UK
  • 2015 Winton Group Leader, the Francis Crick Institute, London, UK