Reliability of panel-based mutational signatures for immune-checkpoint-inhibition efficacy prediction in non-small cell lung cancer.

OBJECTIVES: Mutational signatures (MS) are gaining traction for deriving therapeutic insights for immune checkpoint inhibition (ICI). We asked if MS attributions from comprehensive targeted sequencing assays are reliable enough for predicting ICI efficacy in non-small cell lung cancer (NSCLC). METHODS: Somatic mutations of m = 126 patients were assayed using panel-based sequencing of 523 cancer-related genes. In silico simulations of MS attributions for various panels were performed on a separate dataset of m = 101 whole genome sequenced patients. Non-synonymous mutations were deconvoluted using COSMIC v3.3 signatures and used to test a previously published machine learning classifier. RESULTS: The ICI efficacy predictor performed poorly with an accuracy of 0.51-0.09+0.09, average precision of 0.52-0.11+0.11, and an area under the receiver operating characteristic curve of 0.50-0.09+0.10. Theoretical arguments, experimental data, and in silico simulations pointed to false negative rates (FNR) related to panel size. A secondary effect was observed, where deconvolution of small ensembles of point mutations lead to reconstruction errors and misattributions. CONCLUSION: MS attributions from current targeted panel sequencing are not reliable enough to predict ICI efficacy. We suggest that, for downstream classification tasks in NSCLC, signature attributions be based on whole exome or genome sequencing instead.

Using genomic scars to select immunotherapy beneficiaries in advanced non-small cell lung cancer.

In advanced non-small cell lung cancer (NSCLC), response to immunotherapy is difficult to predict from pre-treatment information. Given the toxicity of immunotherapy and its financial burden on the healthcare system, we set out to identify patients for whom treatment is effective. To this end, we used mutational signatures from DNA mutations in pre-treatment tissue. Single base substitutions, doublet base substitutions, indels, and copy number alteration signatures were analysed in [Formula: see text] patients (the discovery set). We found that tobacco smoking signature (SBS4) and thiopurine chemotherapy exposure-associated signature (SBS87) were linked to durable benefit. Combining both signatures in a machine learning model separated patients with a progression-free survival hazard ratio of 0.40[Formula: see text] on the cross-validated discovery set and 0.24[Formula: see text] on an independent external validation set ([Formula: see text]). This paper demonstrates that the fingerprints of mutagenesis, codified through mutational signatures, select advanced NSCLC patients who may benefit from immunotherapy, thus potentially reducing unnecessary patient burden.

Genome-wide functional element detection using pairwise statistical alignment outperforms multiple genome footprinting techniques.

MOTIVATION: Comparative genomic sequence analysis is a powerful approach for identifying putative functional elements in silico. The availability of full-genome sequences from many vertebrate species has resulted in the development of popular tools, for example, the phastCons software package that search large numbers of genomes to identify conserved elements. While phastCons can analyze many genomes simultaneously, it ignores potentially informative insertion and deletion events and relies on a fixed, precomputed multiple sequence alignment. RESULTS: We have developed a new method, GRAPeFoot, which simultaneously aligns two full genomes and annotates a set of conserved regions exhibiting reduced rates of insertion, deletion and substitution mutations. We tested GRAPeFoot using the human and mouse genomes and compared its performance to a set of phastCons predictions hosted on the UCSC genome browser. Our results demonstrate that despite the use of only two genomes, GRAPeFoot identified constrained elements at rates comparable with phastCons, which analyzed data from 28 vertebrate genomes. This study demonstrates how integrated modelling of substitutions, indels and purifying selection allows a pairwise analysis to exhibit a sensitivity similar to a heuristic analysis of many genomes. AVAILABILITY: The GRAPeFoot software and set of genome-wide functional element predictions are freely available to download online at http://www.stats.ox.ac.uk/ approximately satija/GRAPeFoot/.

Evolution of primate gene expression: drift and corrective sweeps?

Changes in gene expression play an important role in species' evolution. Earlier studies uncovered evidence that the effect of mutations on expression levels within the primate order is skewed, with many small downregulations balanced by fewer but larger upregulations. In addition, brain-expressed genes appeared to show an increased rate of evolution on the branch leading to human. However, the lack of a mathematical model adequately describing the evolution of gene expression precluded the rigorous establishment of these observations. Here, we develop mathematical tools that allow us to revisit these earlier observations in a model-testing and inference framework. We introduce a model for skewed gene-expression evolution within a phylogenetic tree and use a separate model to account for biological or experimental outliers. A Bayesian Markov chain Monte Carlo inference procedure allows us to infer the phylogeny and other evolutionary parameters, while quantifying the confidence in these inferences. Our results support previous observations; in particular, we find strong evidence for a sustained positive skew in the distribution of gene-expression changes in primate evolution. We propose a "corrective sweep" scenario to explain this phenomenon.

A "Long Indel" model for evolutionary sequence alignment.

We present a new probabilistic model of sequence evolution, allowing indels of arbitrary length, and give sequence alignment algorithms for our model. Previously implemented evolutionary models have allowed (at most) single-residue indels or have introduced artifacts such as the existence of indivisible "fragments." We compare our algorithm to these previous methods by applying it to the structural homology dataset HOMSTRAD, evaluating the accuracy of (1) alignments and (2) evolutionary time estimates. With our method, it is possible (for the first time) to integrate probabilistic sequence alignment, with reliability indicators and arbitrary gap penalties, in the same framework as phylogenetic reconstruction. Our alignment algorithm requires that we evaluate the likelihood of any specific path of mutation events in a continuous-time Markov model, with the event times integrated out. To this effect, we introduce a "trajectory likelihood" algorithm (Appendix A). We anticipate that this algorithm will be useful in more general contexts, such as Markov Chain Monte Carlo simulations.

An efficient algorithm for statistical multiple alignment on arbitrary phylogenetic trees.

We present an efficient algorithm for statistical multiple alignment based on the TKF91 model of Thorne, Kishino, and Felsenstein (1991) on an arbitrary k-leaved phylogenetic tree. The existing algorithms use a hidden Markov model approach, which requires at least O( radical 5(k)) states and leads to a time complexity of O(5(k)L(k)), where L is the geometric mean sequence length. Using a combinatorial technique reminiscent of inclusion/exclusion, we are able to sum away the states, thus improving the time complexity to O(2(k)L(k)) and considerably reducing memory requirements. This makes statistical multiple alignment under the TKF91 model a definite practical possibility in the case of a phylogenetic tree with a modest number of leaves.