Data management is the most challenging aspect of building Machine Learning (ML) systems. ML systems can read large volumes of historical data when training models, but inference workloads are more varied, depending on whether it is a batch or online ML system. The feature store for ML has recently emerged as a single data platform for managing ML data throughout the ML lifecycle, from feature engineering to model training to inference. In this paper, we present the Hopsworks feature store for machine learning as a highly available platform for managing feature data with API support for columnar, row-oriented, and similarity search query workloads. We introduce and address challenges solved by the feature stores related to feature reuse, how to organize data transformations, and how to ensure correct and consistent data between feature engineering, model training, and model inference. We present the engineering challenges in building high-performance query services for a feature store and show how Hopsworks outperforms existing cloud feature stores for training and online inference query workloads.

Human papillomavirus (HPV) is an established risk factor for oropharyngeal squamous cell carcinoma (OSCC). The aim was to establish cell lines from HPV-positive tonsil carcinomas to be used for treatment development. Fresh samples from 23 HPV-positive tonsil carcinomas were cultivated in vitro. The established cell line was analyzed for viral characteristics, cell karyotype, TP53 status, and growth capabilities in nude mice. In vitro studies of sensitivities to radiation, cisplatin and cetuximab were performed. After 19 months (eight passages), one cell line, LU-HNSCC-26, was established in vitro and also grew as xenografts. The tumor was from a 48 year old non-smoking man with non-keratinizing, p16 positive tonsil OSCC, stage T2N0M0 with HPV16. It contained 19.5 (CV% 3.7) HPV16 copies/cell (passage 8). The complete HPV16 genome sequence was obtained. Episomal HPV16 was present with an E2/E7 ratio of 1.1 (CV% 2.6). In addition, HPV16 mRNA specific for the intact E2 gene was detected. The viral expression manifested 1.0 (CV% 0.1) E7 mRNA copies per HPV16 genome. The karyotype was determined and the cell line demonstrated wild type TP53. The ID50 for radiation was 0.90 Gy and the IC50 for cisplatin was 0.99 μmol/L. The cell line was inhibited to a maximum of 18% by cetuximab. We established an in vitro tonsil carcinoma cell line containing episomal HPV16. This is an important step towards efficient treatment development.

Basic Local Alignment Search Tool (BLAST) is an essential algorithm that researchers use for sequence alignment analysis. The National Center for Biotechnology Information (NCBI)-BLAST application is the most popular implementation of the BLAST algorithm. It can run on a single multithreading node. However, the volume of nucleotide and protein data is fast growing, making single node insufficient. It is more and more important to develop high-performance computing solutions, which could help researchers to analyze genetic data in a fast and scalable way. This article presents execution of the BLAST algorithm on high performance computing (HPC) clusters and supercomputers in a massively parallel manner using thousands of processors. The Parallel Computing in Java (PCJ) library has been used to implement the optimal splitting up of the input queries, the work distribution, and search management. It is used with the nonmodified NCBI-BLAST package, which is an additional advantage for the users. The result application-PCJ-BLAST-is responsible for reading sequence for comparison, splitting it up and starting multiple NCBI-BLAST executables. Since I/O performance could limit sequence analysis performance, the article contains an investigation of this problem. The obtained results show that using Java and PCJ library it is possible to perform sequence analysis using hundreds of nodes in parallel. We have achieved excellent performance and efficiency and we have significantly reduced the time required for sequence analysis. Our work also proved that PCJ library could be used as an effective tool for fast development of the scalable applications.

Motivation: Next Generation Sequencing (NGS) technology enables identification of microbial genomes from massive amount of human microbiomes more rapidly and cheaper than ever before. However, the traditional sequential genome analysis algorithms, tools, and platforms are inefficient for performing large-scale metagenomic studies on ever-growing sample data volumes. Currently, there is an urgent need for scalable analysis pipelines that enable harnessing all the power of parallel computation in computing clusters and in cloud computing environments. We propose ViraPipe, a scalable metagenome analysis pipeline that is able to analyze thousands of human microbiomes in parallel in tolerable time. The pipeline is tuned for analyzing viral metagenomes and the software is applicable for other metagenomic analyses as well. ViraPipe integrates parallel BWA-MEM read aligner, MegaHit De novo assembler, and BLAST and HMMER3 sequence search tools. We show the scalability of ViraPipe by running experiments on mining virus related genomes from NGS datasets in a distributed Spark computing cluster. Results: ViraPipe analyses 768 human samples in 210 minutes on a Spark computing cluster comprising 23 nodes and 1288 cores in total. The speedup of ViraPipe executed on 23 nodes was 11x compared to the sequential analysis pipeline executed on a single node. The whole process includes parallel decompression, read interleaving, BWA-MEM read alignment, filtering and normalizing of non-human reads, De novo contigs assembling, and searching of sequences with BLAST and HMMER3 tools.

Studies investigating presence of viruses in cancer often analyze case series of cancers, resulting in detection of many viruses that are not etiologically linked to the tumors where they are found. The incidence of virus-associated cancers is greatly increased in immunocompromised individuals. Non-melanoma skin cancer (NMSC) is also greatly increased and a variety of viruses have been detected in NMSC. As immunosuppressed patients often develop multiple independent NMSCs, we reasoned that viruses consistently present in independent tumors might be more likely to be involved in tumorigenesis. We sequenced 8 different NMSCs from 1 patient in comparison to 8 different NMSCs from 8 different patients. Among the latter, 12 different virus sequences were detected, but none in more than 1 tumor each. In contrast, the patient with multiple NMSCs had human papillomavirus type 15 and type 38 present in 6 out of 8 NMSCs.

Most cancer forms known to be caused by viruses are increased among the immunosuppressed, but several cancer forms without established viral etiology are also increased, notably nonmelanoma skin carcinoma (NMSC). We followed all 13,429 solid organ transplantation patients in Sweden for cancer occurrence after transplantation. We requested these tumor specimens and sequenced the first 89 specimens received (62 NMSCs, 27 other cancers). The sequences were analyzed for viruses based on two bioinformatics algorithms (paracel-blast (sensitive for detection of known viruses) and hidden Markov model (HMM; sensitive for distantly related viruses)). Among the 62 NMSCs, the virus family detected in the largest proportion of specimens was Mimiviridae (9/62 NMSCs). The majority of the virus-related reads belonged to Papillomaviridae. The HMM analysis identified 86 additional previously not described viral contigs related to 11 virus families, with reads related to Mimiviridae being the most common (detected in 28/62 NMSCs) with the most prevalent contig (Mimivirus SE906, 1937 bp) detected in 17/62 NMSCs. Among the 27 other cancers, viral sequences were detected in only 5 specimens by blast analysis, compared to in all 27 specimens by HMM (Mimiviridae, Poxviridae, Phycodnaviridae and virus-related sequences yet unclassified to any family). 99% of the virus reads belonged to a single previously not described sequence (Mimivirus SE996, 911 bp). A multitude of viruses is readily detectable in specimens with cancers occurring among the immunosuppressed, with sequences related to Mimiviridae being the most prevalent. Further research would be needed to elucidate the biological significance of the viruses.

This article presents massively parallel execution of the BLAST algorithm on supercomputers and HPC clusters using thousands of processors. Our work is based on the optimal splitting up the set of queries running with the non-modified NCBI-BLAST package for sequence alignment. The work distribution and search management have been implemented in Java using a PCJ (Parallel Computing in Java) library. The PCJ-BLAST package is responsible for reading sequence for comparison, splitting it up and start multiple NCBI-BLAST executables. We also investigated a problem of parallel I/O and thanks to PCJ library we deliver high throughput execution of BLAST. The presented results show that using Java and PCJ library we achieved very good performance and efficiency. In result, we have significantly reduced time required for sequence analysis. We have also proved that PCJ library can be used as an efficient tool for fast development of the scalable applications.

A possible role for infections of the pregnant mother in the development of childhood acute leukemias and lymphomas has been suggested. However, no specific infectious agent has been identified. Offspring of 74,000 mothers who had serum samples taken during pregnancy and stored in a large-scale biobank were followed up to the age of 15 years (750,000 person years) through over-generation linkages between the biobank files, the Swedish national population and cancer registers to identify incident leukemia/lymphoma cases in the offspring. First-trimester sera from mothers of 47 cases and 47 matched controls were retrieved and analyzed using next generation sequencing. Anelloviruses were the most common viruses detected, found in 37/47 cases and in 40/47 controls, respectively (OR: 0.6, 95% CI: 0.2-1.9). None of the detected viruses was associated with leukemia/lymphoma in the offspring. Viremia during pregnancy was common, but no association with leukemia/lymphoma risk in the offspring was found.What's New? To search for prenatal viral infections as a possible cause for acute childhood leukemias and lymphomas, the authors sequenced archival serum samples from 47 pregnant mothers whose children later developed leukemia or lymphoma. Anelloviruseslarge single-stranded DNA viruses with circular genomeswere the most common viruses identified but no statistically significant association with leukemia and lymphoma risk was found when compared to matched controls. The authors point out that larger studies, preferably with serial samples, are necessary to conclusively rule out any risk.

We tested prostatic secretions from men with and without prostate cancer (13 cases and 13 matched controls) or prostatitis (18 cases and 18 matched controls) with metagenomic sequencing. A large number (>200) of viral reads was only detected among four prostate cancer cases (1 patient each positive for Merkel cell polyomavirus, JC polyomavirus and Human Papillomavirus types 89 or 40, respectively). Lower numbers of reads from a large variety of viruses were detected in all patient groups. Our knowledge of the biology of the prostate may be furthered by the fact that DNA viruses are commonly shed from the prostate and can be readily detected by metagenomic sequencing of expressed prostate secretions.

Immunosuppression involves an inability to control virus infections and increased incidence of virus-associated cancers. Some cancers without known viral etiology are also increased, but data on exactly which cancer forms are increased has been inconsistent. To provide a reliable and generalizable estimate, with high statistical power and long follow-up time, we assessed cancer risks using comprehensive, population-based registries in two different countries and from two different immunosuppressed patient groups (solid organ transplant recipients (OTRs) and long-term dialysis patients (LDPs)). National registries in Denmark and Sweden identified 20,804 OTRs and 31,140 LDPs that were followed up using national cancer registries. Standardized incidence ratios (SIR) compared to the general population were estimated. We found highly similar results, both for the two different countries and for the two different immunosuppressed cohorts, namely an increased incidence for the following specific cancer forms: Non-melanoma skin cancer (NMSC), non-Hodgkin's lymphoma and cancers of the lip, kidney, larynx and thyroid. The SIR for overall cancer among OTRs was 3.5 [n = 2,142, 95% CI, 3.4-3.7] in Sweden, 2.9 [n = 1,110, 95% CI, 2.8-3.1] in Denmark and 1.6 [n = 1,713, 95% CI, 1.5-1.6] among LDP. The SIR for NMSC among OTRs was 44.7 [n = 994, 95% CI, 42-47.5] in Sweden and 41.5 [n = 445, 95% CI, 37.8-45.5] in Denmark. The increased SIR for NMSC among LDPs was 5.3 [n = 304, 95% CI, 4.7-5.9]). In summary, an increased SIR for a specific, similar set of cancer forms is consistently found among the immunosuppressed. Conceivable explanations include surveillance bias and immunosuppression-related susceptibility to viral infections.

Scientific Reports 6: Article number: 25235; published online: 28 April 2016; updated: 31 May 2016 In the original version of this Article, Affiliations 1 and 2 were not listed in the correct order. The correct affiliations are listed below: Affiliation 1 Karolinska Institutet, Department of Laboratory Medicine, Stockholm, 141 86, Sweden.

Established Human Papillomavirus (HPV) types, up to HPV202, belong to 49 species in five genera. International standardization in classification and quality standards for HPV type designation and detection is ensured by the International HPV Reference Center. The center i) receives clones of potentially novel HPV types, re-clones and re-sequences them. If confirmed, an HPV type number is assigned and posted on www.hpvcenter.se. ii) distributes reference clone samples, for academic research, under Material Transfer Agreements agreed with the originator. iii) provides preliminary checking of whether new sequences represent novel types iv) issues international proficiency panels for HPV genotyping. The rate of HPV type discovery is increasing, probably because of metagenomic sequencing. γ-genus today contains 79HPV types and 27 species, surpassing ∝ and β genera with 65 and 51HPV types, respectively. Regular issuing of proficiency panels based on HPV reference clones has resulted in global improvement of HPV genotyping services.

Condylomata acuminata is caused by human papillomavirus (HPV). PCR with consensus primers will typically detect HPV in >96% of condylomata. Metagenomic sequencing has found that some "HPV-negative" condylomata do indeed contain HPV. We wished to perform a renewed evaluation of the "HPV-negative" condylomata using deeper metagenomics sequencing. Sequencing of whole genome amplified DNA from 40 apparently "HPV-negative" condylomata detected HPV in 37/40 specimens. We found 75 different HPV types, out of which 43 represented novel putative HPV types. Three types were cloned and established as HPV types 200, 201 and 202. Molluscum contagiosum virus was detected in 24 of the 40 samples. In summary, deep sequencing enables detection of HPV in almost all condylomata. "HPV-negative" condylomata might largely be explained by clinical misdiagnosis or the presence of viral variants, distantly related HPV types and/or low viral loads.

Papillomaviruses (PV) are a remarkably heterogeneous family of small DNA viruses that infect a wide variety of vertebrate species and are aetiologically linked with the development of various neoplastic changes of the skin and mucosal epithelia. Based on nucleotide similarity, PVs are hierarchically classified into genera, species and types. Novel human PV (HPV) types are given a unique number only after the whole genome has been cloned and deposited with the International HPV Reference Center. As of 9 March 2015, 200 different HPV types, belonging to 49 species, had been recognized by the International HPV Reference Center. In addition, 131 animal PV types identified from 66 different animal species exist. Recent advances in molecular techniques have resulted in an explosive increase in the identification of novel HPV types and novel subgenomic HPV sequences in the last few years. Among PV genera, the γ-PV genus has been growing most rapidly in recent years with 80 completely sequenced HPV types, followed by α-PV and β-PV genera that have 65 and 51 recognized HPV types, respectively. We reviewed in detail the contemporary molecular methods most often used for identification and characterization of novel PV types, including PCR, rolling circle amplification and next-generation sequencing. Furthermore, we present a short overview of 12 and 10 novel HPV types recently identified in Sweden and Slovenia, respectively. Finally, an update on the International Human Papillomavirus Reference Center is provided.

Patients with recurrent respiratory papillomatosis (RRP) in Norway treated between 1987 and 2009 were recruited to this cohort study. They were followed from disease onset and data recorded until January 2012. Here, we describe the distribution of human papillomavirus (HPV) genotypes, the prevalence of multiple HPV infections, and the risk of high-grade laryngeal neoplasia and respiratory tract invasive carcinoma in a large cohort of patients with RRP. We also examined whether HPV genotype, gender, age or clinical course are risk factors for this development. Clinical records and histological specimens were reviewed. Using formalin-fixed paraffin-embedded biopsies, HPV genotyping were performed by quantitative polymerase chain reaction assays identifying 15 HPV types. HPV-negative specimens were analyzed by metagenomic sequencing. Paraffin blocks were available in 224/238 patients. The DNA quality was approved in 221/224 cases. HPV DNA was detected in 207/221 patients and all were HPV 6 or HPV 11 positive, comprising HPV 6 in 133/207, HPV 11 in 40/207 cases and HPV 6/11 in 15/207 cases. Co-infection with one or two high-risk HPV types together with HPV 6 or HPV 11 was present in 19/207 patients. Metagenomic sequencing of 14 HPV-negative specimens revealed HPV 8 in one case. In total, 39/221 patients developed high-grade laryngeal neoplasia. 8/221 patients developed carcinoma of the respiratory tract (six patients with laryngeal carcinoma and two patients with lung carcinoma). High-grade laryngeal neoplasias were found more frequently in HPV-negative versus HPV-positive patients, (RR = 2.35, 95% CI 1.1, 4.99), as well as respiratory tract carcinomas (RR = 48, 95% CI 10.72, 214.91). In summary, the majority of RRP were associated with HPV 6 and/or 11. HPV-negative RRP biopsies occurred more frequently in adult-onset patients, and were associated with an increased risk of laryngeal neoplasia and carcinoma in the respiratory tract.

Most viruses in human skin are known to be human papillomaviruses (HPVs). Previous sequencing of skin samples has identified 273 different cutaneous HPV types, including 47 previously unknown types. In the present study, we wished to extend prior studies using deeper sequencing. This deeper sequencing without prior PCR of a pool of 142 whole genome amplified skin lesions identified 23 known HPV types, 3 novel putative HPV types and 4 non-HPV viruses. The complete sequence was obtained for one of the known putative types and almost the complete sequence was obtained for one of the novel putative types. In addition, sequencing of amplimers from HPV consensus PCR of 326 skin lesions detected 385 different HPV types, including 226 previously unknown putative types. In conclusion, metagenomic deep sequencing of human skin samples identified no less than 396 different HPV types in human skin, out of which 229 putative HPV types were previously unknown.

Non‐melanoma skin cancers commonly contain Human Papillomavirus (HPV), but the types found have varied depending on the polymerase chain reaction (PCR) primer systems used. Whole genome amplified DNA (not amplified by any specific PCR primers) from 91 skin lesions [41 squamous cell skin carcinomas (SCCs), 8 keratoacanthomas, 22 actinic keratoses, 3 basal cell carcinomas and 17 SCCs in situ ] were sequenced. All samples were sequenced both at 160 Mb and 1.8 Gb sequencing depth per sample. The sequences from 10 different HPVs in 47/91 specimens were found. Sequences represented four established HPV types (HPV types 16, 22, 120, 124), two previously known putative types (present in GenBank) and four previously unknown HPV sequences (new putative types). The most commonly detected virus was cloned, sequenced and designated as HPV197. Type‐specific real‐time PCR detected HPV197 in 34/91 specimens. For comparison, a pool of the same samples after general primer PCR amplification was also sequenced. This revealed 40 different HPVs, but only two HPV types were detected both with sequencing without prior PCR and with sequencing PCR amplicons, suggesting that sequencing without prior PCR gives a more unbiased representation of the HPVs present. In summary, it was found that HPV can be sequenced from most skin disease specimens and HPV197 appeared to be the most commonly present virus.