CONIPHER: a computational framework for scalable phylogenetic reconstruction with error correction

Abstract

Intra-tumour heterogeneity provides the fuel for the evolution and selection of subclonal tumour cell populations. However, accurate inference of tumour subclonal architecture and reconstruction of tumour evolutionary history from bulk DNA sequencing data remains challenging. Sequencing and alignment artefacts cannot be distinguished from real cancer somatic mutations and errors in the identification of copy number alterations or complex evolutionary events (e.g. mutation losses) affect the estimated cellular prevalence of mutations, leading to errors in mutation clustering and phylogenetic reconstructions. In this paper we present a new computational framework, CONIPHER (COrrecting Noise In PHylogenetic Evaluation and Reconstruction), that accurately infers subclonal structure and phylogenetic relationships from multi-sample tumour sequencing, accounting for both copy number alterations and mutation errors. CONIPHER outperforms similar methods on simulated datasets, and in particular scales to a large number of tumour samples and clones. As such, CONIPHER enables automated phylogenetic analysis which can be effectively applied to large sequencing datasets generated with different technologies.

The evolution of lung cancer and impact of subclonal selection in TRACERx

“…To decipher the timing of somatic events in each tumour, we attempted to construct tumour phylogenetic trees from the identified somatic alterations. In total, 1,553 freshly frozen surgically excised tumour regions were analysed, excluding 1 region that harboured a collision between 2 genomically distinct tumours (Methods) 11 . These included 1,515 primary tumour and 38 lymph node regions sampled at surgery.…”

“…To reconstruct tumour phylogenetic trees of each tumour from the identified somatic mutations, we developed a new computational method called CONIPHER 11 (correcting noise in phylogenetic evaluation and reconstruction) to address three key challenges in phylogenetic reconstruction: (1) scaling to a high number of primary tumour and metastasis regions per patient; (2) correcting for complex evolutionary events, including mutation losses 25 ; (3) removing biologically improbable clusters that are either driven by subclonal copy number or are not biologically compatible with the inferred evolutionary tree. CONIPHER includes three key steps, which we describe in brief below.…”

“…CONIPHER includes three key steps, which we describe in brief below. We report the full details in a companion paper 11 .…”

“…We define the phylogenetic cancer cell fraction (PhyloCCF) of each mutation to be the fraction of cancer cells that either carry a SNV or which are estimated to have carried a SNV before a subclonal copy number loss of alleles carrying the SNV 11,25 . This statistic is estimated by transforming the VAF by the expected mutation copy number and the tumour purity to compute the standard cancer cell fraction (CCF) metric 25,30,61 , and correcting for subclonal copy number alterations, as previously described 3 .…”

The evolution of non-small cell lung cancer metastases in TRACERx

“…Phylogenetic trees were constructed for each case using our tool CONIPHER 4 , and the timing of metastatic divergence was estimated (defined as when the metastatic clone first existed, rather than when the cells migrated from the primary tumour; Methods). We defined two broad categories of metastatic divergence timing: early or late (Fig.…”