Erich Allen Peterson

Clonal selection and double-hit events involving tumor suppressor genes underlie relapse in myeloma.

To elucidate the mechanisms underlying relapse from chemotherapy in multiple myeloma, we performed a longitudinal study of 33 patients entered into Total Therapy protocols investigating them using gene expression profiling, high-resolution copy number arrays, and whole-exome sequencing. The study illustrates the mechanistic importance of acquired mutations in known myeloma driver genes and the critical nature of biallelic inactivation events affecting tumor suppressor genes, especially TP53, the end result being resistance to apoptosis and increased proliferation rates, which drive relapse by Darwinian-type clonal evolution. The number of copy number aberration changes and biallelic inactivation of tumor suppressor genes was increased in GEP70 high risk, consistent with genomic instability being a key feature of high risk. In conclusion, the study highlights the impact of acquired genetic events, which enhance the evolutionary fitness level of myeloma-propagating cells to survive multiagent chemotherapy and to result in relapse.

Mining probabilistic frequent closed itemsets in uncertain databases

This paper defines probabilistic support and probabilistic frequent closed itemsets in uncertain databases for the first time. It also proposes a probabilistic frequent closed itemset mining (PFCIM) algorithm to mine probabilistic frequent closed itemsets from uncertain databases.

Towards the integration, annotation and association of historical microarray experiments with RNA-seq.

BackgroundTranscriptome analysis by microarrays has produced important advances in biomedicine. For instance in multiple myeloma (MM), microarray approaches led to the development of an effective disease subtyping via cluster assignment, and a 70 gene risk score. Both enabled an improved molecular understanding of MM, and have provided prognostic information for the purposes of clinical management. Many researchers are now transitioning to Next Generation Sequencing (NGS) approaches and RNA-seq in particular, due to its discovery-based nature, improved sensitivity, and dynamic range. Additionally, RNA-seq allows for the analysis of gene isoforms, splice variants, and novel gene fusions. Given the voluminous amounts of historical microarray data, there is now a need to associate and integrate microarray and RNA-seq data via advanced bioinformatic approaches.MethodsCustom software was developed following a model-view-controller (MVC) approach to integrate Affymetrix probe set-IDs, and gene annotation information from a variety of sources. The tool/approach employs an assortment of strategies to integrate, cross reference, and associate microarray and RNA-seq datasets.ResultsOutput from a variety of transcriptome reconstruction and quantitation tools (e.g., Cufflinks) can be directly integrated, and/or associated with Affymetrix probe set data, as well as necessary gene identifiers and/or symbols from a diversity of sources. Strategies are employed to maximize the annotation and cross referencing process. Custom gene sets (e.g., MM 70 risk score (GEP-70)) can be specified, and the tool can be directly assimilated into an RNA-seq pipeline.ConclusionA novel bioinformatic approach to aid in the facilitation of both annotation and association of historic microarray data, in conjunction with richer RNA-seq data, is now assisting with the study of MM cancer biology.

Mining frequent sequential patterns with first-occurrence forests

In this paper, a new pattern-growth algorithm is presented to mine frequent sequential patterns using First-Occurrence Forests (FOF). This algorithm uses a simple list of pointers to the first-occurrences of a symbol in the aggregate tree [1], as the basic data structure for database representation, and does not rebuild aggregate trees for projection databases. The experimental evaluation shows that our new FOF mining algorithm outperforms the PLWAP-tree mining algorithm [2] and the FLWAP-tree mining algorithm [3], both in the mining time and the amount of memory used.

Bi-allelic inactivation is more prevalent at relapse in multiple myeloma, identifying RB1 as an independent prognostic marker

The purpose of this study is to identify prognostic markers and treatment targets using a clinically certified sequencing panel in multiple myeloma. We performed targeted sequencing of 578 individuals with plasma cell neoplasms using the FoundationOne Heme panel and identified clinically relevant abnormalities and novel prognostic markers. Mutational burden was associated with maf and proliferation gene expression groups, and a high-mutational burden was associated with a poor prognosis. We identified homozygous deletions that were present in multiple myeloma within key genes, including CDKN2C, RB1, TRAF3, BIRC3 and TP53, and that bi-allelic inactivation was significantly enriched at relapse. Alterations in CDKN2C, TP53, RB1 and the t(4;14) were associated with poor prognosis. Alterations in RB1 were predominantly homozygous deletions and were associated with relapse and a poor prognosis which was independent of other genetic markers, including t(4;14), after multivariate analysis. Bi-allelic inactivation of key tumor suppressor genes in myeloma was enriched at relapse, especially in RB1, CDKN2C and TP53 where they have prognostic significance.

Fast approximation of probabilistic frequent closed itemsets

In recent years, the concept of and algorithm for mining probabilistic frequent itemsets (PFIs) in uncertain databases, based on possible worlds semantics and a dynamic programming approach for frequency calculations, has been proposed. The frequentness of a given itemset in this scheme can be characterized by the Poisson binomial distribution. Further and more recently, others have extended those concepts to mine for probabilistic frequent closed itemsets (PFCIs), in an attempt to reduce the number and redundancy of output. In addition, work has been done to accelerate the computation of PFIs through approximation, to mine approximate probabilistic frequent itemsets (A-PFIs), based on the fact that the Poisson distribution can closely approximate the Poisson binomial distribution---especially when the size of the database is large. In this paper, we introduce the concept of and an algorithm for mining approximate probabilistic frequent closed itemsets (A-PFCIs). A new mining algorithm for mining such concepts is introduced and called A-PFCIM. It is shown through an experimental evaluation that mining for A-PFCIs can be orders of magnitude faster than mining for traditional PFCIs.

Revealing the inherent heterogeneity of human malignancies by variant consensus strategies coupled with cancer clonal analysis

Tumors are heterogeneous in composition. They are composed of cancer cells proper, along with stromal elements that collectively form a microenvironment, all of which are necessary to nurture the malignant process. In addition, many of the stromal cells are modified to support the unique needs of the malignant state. Tumors are composed of a variety of clones or subpopulations of cancer cells, which may differ in karyotype, growth rate, expression of cell surface markers, sensitivity to therapeutics, etc. New tools and methods to provide an improved understanding of tumor clonal architecture are needed to guide therapy.The subclonal structure and transcription status of underlying somatic mutations reveal the trajectory of tumor progression in patients with cancer. Approaching the analysis of tumors to reveal clonal complexity in a quantitative manner should facilitate better characterization and therapeutic assignments. The challenge is the interpretation of massive amounts of data from next generation sequencing (NGS) experiments to find what is truly meaningful for improving the understanding of basic cancer biology, as well as therapeutic assignments and outcomes. To meet this need, a methodology named CloneViz was developed and utilized for the identification of serial clonal mutations.Whole exome sequencing (WES) on an Illumina HiSeq 2500 was performed on paired tumor and normal samples from a Multiple Myeloma (MM) patient at presentation, then first and second relapse. Following alignment, a consensus strategy for variant selection was employed along with computational linkage to a formal tumor clonality analysis based on visualization and quantitative methods.

Leveraging the new with the old: providing a framework for the integration of historic microarray studies with next generation sequencing

Next Generation Sequencing (NGS) methods are rapidly providing remarkable advances in our ability to study the molecular profiles of human cancers. However, the scientific discovery offered by NGS also includes challenges concerning the interpretation of large and non-trivial experimental results. This task is potentially further complicated when a multitude of molecular profiling modalities are available, with the goal of a more integrative and comprehensive analysis of the cancer biology.Microarray transcriptome analyses have resulted in important advances in both the scientific and clinical domains of biomedicine. Importantly, as technology advances, it is critical to leverage what has been gained from historic approaches (e.g., microarrays) with new approaches (NGS). In this regard, necessity dictated a need to utilize and leverage the many years of historical microarray data with new NGS approaches. This is especially important since NGS approaches are now entering clinical medicine. For instance, NGS-based comprehensive analysis of certain cancers has already helped to uncover specific mutations that contribute to the malignant process, identify new therapeutic targets, and improve opportunities for choosing the best treatment for an individual patient.A suite of custom software tools have been developed to rapidly integrate, explore, discover and validate molecular profiling data from the NGS modalities of Whole Exome Sequencing (WES) and RNA-seq with each other, as well as with historical microarray and salient clinical datasets. Importantly, our approach is independent of any particular type of NGS suite(s) or cancer types. This novel bioinformatic framework is now assisting with the scientific and clinical management of patients with multiple myeloma.

Liquid biopsy and its role in an advanced clinical trial for lung cancer

Liquid biopsy methodologies, for the purpose of plasma genotyping of cell-free DNA (cfDNA) of solid tumors, are a new class of novel molecular assays. Such assays are rapidly entering the clinical sphere of research-based monitoring in translational oncology, especially for thoracic malignancies. Potential applications for these blood-based cfDNA assays include: (i) initial diagnosis, (ii) response to therapy and follow-up, (iii) tumor evolution, and (iv) minimal residual disease evaluation. Precision medicine will benefit from cutting-edge molecular diagnostics, especially regarding treatment decisions in the adjuvant setting, where avoiding over-treatment and unnecessary toxicity are paramount. The use of innovative genetic analysis techniques on individual patient tumor samples is being pursued in several advanced clinical trials. Rather than using a categorical treatment plan, the next critical step of therapeutic decision making is providing the “right” cancer therapy for an individual patient, including correct dose and timeframe based on the molecular analysis of the tumor in question. Per the 21st Century Cures Act, innovative clinical trials are integral for biomarker and drug development. This will include advanced clinical trials utilizing: (i) innovative assays, (ii) molecular profiling with cutting-edge bioinformatics, and (iii) clinically relevant animal or tissue models. In this paper, a mini-review addresses state-of-the-art liquid biopsy approaches. Additionally, an on-going advanced clinical trial for lung cancer with novelty through synergizing liquid biopsies, co-clinical trials, and advanced bioinformatics is also presented. Impact statement Liquid biopsy technology is providing a new source for cancer biomarkers, and adds new dimensions in advanced clinical trials. Utilizing a non-invasive routine blood draw, the liquid biopsy provides abilities to address perplexing issues of tumor tissue heterogeneity by identifying mutations in both primary and metastatic lesions. Regarding the assessment of response to cancer therapy, the liquid biopsy is not ready to replace medical imaging, but adds critical new information; for instance, through a temporal assessment of quantitative circulating tumor DNA (ctDNA) assay results, and importantly, the ability to monitor for signs of resistance, via emerging clones. Adjuvant therapy may soon be considered based on a quantitative cfDNA assay. As sensitivity and specificity of the technology continue to progress, cancer screening and prevention will improve and save countless lives by finding the cancer early, so that a routine surgery may be all that is required for a definitive cure.

Mining probabilistic association rules from uncertain databases with pruning

In this paper, we rigorously define the problem of mining probabilistic association rules from uncertain databases. We further analyze the probability distribution space of a candidate probabilistic association rule, and propose an efficient mining algorithm with pruning to find all probabilistic association rules from uncertain databases.