Purpose Painful peripheral neuropathy is a frequent toxicity associated with bortezomib therapy. This study aimed to identify loci that affect susceptibility to this toxicity. Experimental Design A genome-wide association study (GWAS) of 370,605 SNPs was performed to identify risk variants for developing severe bortezomib-induced peripheral neuropathy (BiPN) in 469 multiple myeloma (MM) patients who received bortezomib-dexamethasone therapy prior to autologous stem-cell in randomized clinical trials of the Intergroupe Francophone du Myelome (IFM) and findings were replicated in 114 MM patients of the HOVON-65/GMMG-HD4 clinical trial. Results A single SNP in the PKNOX1 gene was associated with BiPN in the exploratory cohort (rs2839629; OR, 1.89, 95% CI: [1.45-2.44]; P = 7.6 × 10−6) and in the replication cohort (OR, 2.04; 95% CI = [1.11-3.33]; P= 8.3 × 10−3). In addition, rs2839629 is in strong linkage disequilibrium (r2 = 0.87) with rs915854, located in the intergenic region between PKNOX1 and CBS. Expression quantitative trait loci mapping showed that both rs2839629 and rs915854 genotypes impact PKNOX1 expression in nerve tissue while rs2839629 affects CBS expression in skin and blood. Conclusions The use of GWAS in MM pharmacogenomics has identified a novel candidate genetic locus mapping to PKNOX1 and in the immediate vicinity of CBS at 21q22.3 associated with the severe bortezomib-induced toxicity. The proximity of these two genes involved in neurologic pain whose tissue-specific expression is modified by the two variants provides new targets for neuro-protective strategies.

We present a novel hybrid algorithm for Bayesian network structure learning, called H2PC. It first reconstructs the skeleton of a Bayesian network and then performs a Bayesian-scoring greedy hill-climbing search to orient the edges. The algorithm is based on divide-and-conquer constraint-based subroutines to learn the local structure around a target variable. We conduct two series of experimental comparisons of H2PC against Max-Min Hill-Climbing (MMHC), which is currently the most powerful state-of-the-art algorithm for Bayesian network structure learning. First, we use eight well-known Bayesian network benchmarks with various data sizes to assess the quality of the learned structure returned by the algorithms. Our extensive experiments show that H2PC outperforms MMHC in terms of goodness of fit to new data and quality of the network structure with respect to the true dependence structure of the data. Second, we investigate H2PC's ability to solve the multi-label learning problem. We provide theoretical results to characterize and identify graphically the so-called minimal label powersets that appear as irreducible factors in the joint distribution under the faithfulness condition. The multi-label learning problem is then decomposed into a series of multi-class classification problems, where each multi-class variable encodes a label powerset. H2PC is shown to compare favorably to MMHC in terms of global classification accuracy over ten multi-label data sets covering different application domains. Overall, our experiments support the conclusions that local structural learning with H2PC in the form of local neighborhood induction is a theoretically well-motivated and empirically effective learning framework that is well suited to multi-label learning. The source code (in R) of H2PC as well as all data sets used for the empirical tests are publicly available.