Search

Contains fulltext : 283142.pdf (Publisher’s version ) (Open Access) Infants developing necrotizing enterocolitis (NEC) have a different metabolomic profile compared to controls. The potential of specific metabolomics, i.e. amino acids and amino alcohols (AAA), as early diagnostic biomarkers for NEC is largely unexplored. In this multicenter prospective case-control study, longitudinally collected fecal samples from preterm infants (born

Colorectal cancer (CRC) is the second leading cause of cancer death in the western world. The majority of CRCs, which develop from adenoma precursor lesions, show gain of chromosome arm 20q, where BCL2L1 is located. BCL2L1 is an important apoptosis regulating gene that codes for both an anti-apoptotic (Bcl-x

We develop an inference framework for the difference in errors between 2 prediction procedures. The 2 procedures may differ in any aspect and possibly utilize different sets of covariates. We apply training and testing on the same data set, which is accommodated by sample splitting. For each split, both procedures predict the response of the same samples, which results in paired residuals to which a signed-rank test is applied. Multiple splits result in multiple p-values. The median p-value and the mean inverse normal transformed p-value are proposed as summary (test) statistics, for which bounds on the overall type I error rate under a variety of assumptions are proven. A simulation study is performed to check type I error control of the least conservative bound. Moreover, it confirms superior power of our method with respect to a one-split approach. Our inference framework is applied to genomic survival data sets to study 2 issues: compare lasso and ridge regression and decide upon use of both methylation and gene expression markers or the latter only. The framework easily accommodates any prediction paradigm and allows comparing any 2, possibly nonmodel-based, prediction procedures.

Background: Multiplex Ligation dependent Probe Amplification (MLPA) is a rapid, simple, reliable and customized method for detection of copy number changes of individual genes at a high resolution and allows for high throughput analysis. This technique is typically applied for studying specific genes in large sample series. The large amount of data, dissimilarities in PCR efficiency among the different probe amplification products, and sample-to-sample variation pose a challenge to data analysis and interpretation. We therefore set out to develop an MLPA data analysis strategy and tool that is simple to use, while still taking into account the above-mentioned sources of variation. Materials and methods: MLPAnalyzer was developed in Visual Basic for Applications, and can accept a large number of file formats directly from capillary sequence systems. Sizes of all MLPA probe signals are determined and filtered, quality control steps are performed, and variation in peak intensity related to size is corrected for. DNA copy number ratios of test samples are computed, displayed in a table view and a set of comprehensive figures is generated. To validate this approach, MLPA reactions were performed using a dedicated MLPA mix on 6 different colorectal cancer cell lines. The generated data were normalized using our program and results were compared to previously performed array-CGH results using both statistical methods and visual examination. Results and discussion: Visual examination of bar graphs and direct ratios for both techniques showed very similar results, while the average Pearson moment correlation over all MLPA probes was found to be 0.42. Our results thus show that automated MLPA data processing following our suggested strategy may be of significant use, especially when handling large MLPA data sets, when samples are of different quality, or interpretation of MLPA electropherograms is too complex. It remains, however, important to recognize that automated MLPA data processing may only be successful when a dedicated experimental setup is also considered. © 2008 - IOS Press and the authors. All rights reserved.

Clinicians often research complex traits in which many variables may be involved in the process underlying the disease. Nowadays, with the advent of DNA sequencing techniques, clinical studies regularly research high‐dimensional data, in which the number of variables even far exceeds the number of samples. Generic research goals are to predict certain outcomes or to select variables, using the high-dimensional data possibly in addition to clinical variables like age and sex. Examples in cancer genomics are to diagnose cancer, classify cancer type and predict survival time based on gene expression data, or to find few genes that may predict these outcomes well. While the human genome contains around 20.000 genes, clinical studies usually only include measurements for around 100 patients. This limited amount of information makes it hard to find the “right” selection or combination of variables from the vast space of options. Luckily, more prior information on the variables is often available in the form of complementary data, or co-data, e.g. from public repositories or derived from domain knowledge. Co‐data may vary in type. Genes, for example, may be grouped in non‐overlapping groups for chromosomes, overlapping groups for pathways, hierarchical groups for gene ontology or assigned a continuous summary statistic derived from a similar study. We would like to learn from co‐data to improve prediction and variable selection. This dissertation presents three statistical methods and software for co‐data learning. We consider co‐data learnt penalised generalised linear and Cox survival models for the outcome. The penalties on the variables are informed by the co‐data, such that variables for more important co‐data are penalised less. For example, some biological functions may be more important than others, such that genes corresponding to these biological functions are ideally penalised relatively less. The penalty parameters are related to prior parameters for a prior distribution on the variables, which are estimated with an empirical Bayes approach. The presented methods differ in the type of co‐data and penalty or prior that may be used.

Colorectal cancer (CRC) is the second most frequent cause of cancer-related death worldwide. Detecting CRC in a premalignant or early stage can improve survival of patients. When the CRC is metastasized (mCRC) to other organs patients are offered systemic treatment to prolong survival. However, not all patients respond to the available treatment while they do suffer from treatment related toxicities. In this thesis we studied the effects of CRC screening programmes on incidence and mortality. In addition, we studied tumor genomics and clinicopathological factors for the prediction of response to systemic therapy in patients with mCRC. The first chapter of this thesis described the effects of CRC screening on incidence and survival of patients with CRC. It was expected that screening for CRC would result in the detection of 1600 additional stage I and II CRCs per year in the first few years after its introduction in The Netherlands. This increase in detected stage I and II resulted in a decrease in the proportion of CRCs diagnosed at stages III or IV from 47% to 20%. Studies described in the next chapters aimed to improve treatment of patients with mCRC and to reduce unnecessary treatment related toxicity by patient selection for systemic therapy. Patient selection can be done with predictive biomarkers based on tumor genomics. A large number of genomic tumor characteristics can be analysed. This high dimensional data makes statistics challenging, as corrections for multiple testing are crucial to avoid overoptimistic results. Moreover, stable estimation of parameters is crucial for determining reproducible biomarkers. In chapter 2 we described a method with multi-parameter shrinkage options to overcome these statistical challenges. In chapter 3 we demonstrated that the miRNA expression profiles of metastases closely resemble that of their corresponding primary CRCs (pCRC). Only 8 (0.5%) of the 1714 miRNAs were significantly different expressed between pCRC and their matched metastases. Based on these results, we expected that miRNA expression profiles of primary tumors and metastases may be of similar predictive value for predicting prognosis or treatment response for patients with mCRC. In chapter 4 we analysed the miRNA expression levels of metastasised colorectal tumors in addition to known predictive clinocopathological factors. This study demonstrated that expression levels of miR-17-5p, miR-20a-5p, miR-30a-5p, miR-92a-3p, miR-92b-3p and miR-98-5p in combination with age, tumor differentiation, adjuvant therapy and type of systemic treatment, were predictive for clinical benefit (response and stable disease (SD)) of first-line chemotherapy in the training cohort with an AUC of 0.78. In the validation cohort the addition of the six miRNA signature to the four clinicopathological factors demonstrated a significant increased AUC for predicting treatment response versus those with SD from 0.79 to 0.90. However, our six miRNA signature did not add predictive value to the four selected clinicopathological factors for separating patients with PD from those with SD or response. Besides miRNAs, other readouts of tumor genomics may be used to improve sensitivity and specificity. In chapter 5 we analysed tumor copy number aberrations, microsatellite instability and known cancer related mutations for their predictive value for treatment response. Based on baseline clinicopathological factors, response to first and second line treatment in patients with mCRC could be predicted with an AUC of 0.73 and 0.69 respectively. Unfortunately, these prediction characteristics could not be improved by the addition of mutation status and copy number aberrations. The addition of clinicopathological factors acquired during first line treatment significantly increased the performance to predict response to second line treatment (AUC of 0.75 (p = 0.04)). Again, this was not sufficient to guide treatment decisions.

Chromosomal copy number aberrations are one of the main mechanisms that give rise to the proliferative capabilities of cancer cells. These aberrations can be quantified with technologies that generate measurements genome-wide and with high resolution. Hence, they produce vast amounts of data, which requires tailored bioinformatic solutions for analysis and management. Two such high-resolution and genome-wide technologies are DNA microarrays, which are successively replaced by next-generation sequencing approaches. This dissertation describes three novel bioinformatic solutions for copy number analysis in cancer with these technologies. CanGEM is a publicly-accessible database solution for storage of raw and processed copy number data from cancer research experiments. The contents of the database can be queried based on clinical and copy number data. Clinical data is collected using appropriate controlled vocabularies. Copy number data is collected as raw microarray data and automated analysis identifies the locations of chromosomal aberrations. In order to allow integration of data measured with different microarray platforms, a copy number status is derived for every known human gene. CGHpower is a statistical power calculator for copy number experiments that compare two groups. It estimates genome complexity of a cancer type in question from a pilot data set of the sample series, and assesses the number of samples required to satisfy statistical requirements. It can be used either in the planning stages of experiments, including as a justification in grant applications, or to verify whether sufficient samples were included in past experiments. Performance of this bioinformatic solution is evaluated with real and simulated data sets. QDNAseq is a preprocessing solution to detect copy number aberrations from shallow whole-genome next-generation sequencing data. It corrects the observed sequencing coverage for known systematic biases and allows filtering of spurious regions in the genome. A new list of such problematic regions is derived from public data generated by the 1000 Genomes Project. Performance of the solution is evaluated relative to other similar published solutions and DNA microarrays, and also compared to theoretical statistical expectations. An application of the QDNAseq method is also presented in a translational research project with the aim to identify copy number aberrations in tumors of patients with low-grade glioma. Aberrations identified by shallow whole-genome next-generation sequencing and QDNAseq are used to evaluate associations with patient survival, and also to assess intratumoral heterogeneity and temporal evolution of these tumors. A loss in chromosome 10q is identified to be associated with poor prognosis, and the finding validated in two independent data sets. From the assessment of intratumoral heterogeneity and temporal tumor evolution, the well-characterized co-deletion of 1p/19q is found to be the only chromosomal aberration that is consistently present or absent across the entire tumor and possible future recurrences. This is compatible with the present view of its role as an early event in the development of these tumors. The text concludes with a discussion of lessons learned from the development process and application of the three described bioinformatic solutions. Better awareness of and adherence to established best practices from the software development field would have been useful, and together with more careful consideration of implementation decisions could have resulted… Kromosomaaliset kopiolukupoikkeamat ovat eräs tärkeimmistä mekanismeista syövän synnyssä. Yhden äidiltä ja yhden isältä perityn geenikopion sijaan osa perimästä voi olla monistunut useammaksi kopioksi, ja joidenkin osien kohdalla yksi tai molemmat kopiot voivat olla hävinneet. Kopiolukupoikkeamien todentamiseen käytetään genominlaajuisia tekniikoita, joilla on tarkka erotuskyky. Ne tuottavat suuria tietomääriä, joiden analysointi ja käsittely vaativat räätälöityjä bioinformaattisia menetelmiä. Tekniikoihin sisältyvät DNA-mikrolevyt sekä ne käytännössä jo syrjäyttäneet uuden sukupolven sekvensointimenetelmät. Tässä väitöskirjassa kuvataan kolme uutta bioinformaattista ohjelmistoa kopiolukupoikkeamien analysointiin syöpänäytteistä näillä tekniikoilla. CanGEM on julkinen tietokanta raa'an ja prosessoidun mikrolevyaineiston keräämiseen yksittäisistä syöpätutkimuksista. Tietokannan sisältöön voi tehdä hakuja kliinisten muuttujien tai kopiolukupoikkeamien perusteella. Kliinisten muuttujien tallennukseen käytetään asianmukaisia luokittelujärjestelmiä. Kopiolukuaineisto kerätään raakoina mikrolevymittauksina, joista kopiolukupoikkeamat tunnistetaan algoritmisesti. Jotta eri mikrolevyalustoilla mitatun tiedon yhdistäminen olisi mahdollista, kopioluku määritetään erikseen jokaiselle tunnetulle ihmisen geenille. CGHpower on menetelmä tilastollisten voima-analyysien tekemiseen kahta ryhmää vertailevista kopiolukututkimuksista. Aineiston kopiolukupoikkeamien monimutkaisuus arvioidaan koe-erästä näytteitä ja määritetään tilastollisten vaatimusten edellyttämä otoskoko. Menetelmää voidaan käyttää joko tutkimusten suunnitteluvaiheessa, mm. rahoitushakemusten tukena, tai arvioimaan onko jo tehdyissä kokeissa käytetty riittävää määrää näytteitä. Suorituskyky mitataan sekä todellisilla että simuloiduilla aineistoilla. QDNAseq on esikäsittelymenetelmä kopiolukupoikkeamien tunnistamiseen matalalla lukupeitolla ja genominlaajuisesti tuotetusta uuden sukupolven sekvensointiaineistosta. Se korjaa havaittua lukupeittoa tunnettujen vinoumalähteiden osalta ja mahdollistaa kopiolukuanalyyseille ongelmallisten perimän osien suodattamisen jatkokäsittelystä. Näistä ongelmallisista alueista kuvataan uusi luettelo, joka on johdettu 1000 Genomes -projektin julkaisemasta aineistosta. Menetelmän suorituskykyä arvioidaan verrattuna muihin vastaaviin julkaistuihin menetelmiin ja DNA-mikrolevyihin, sekä suhteessa teoreettisiin tilastollisiin odotuksiin. Itse menetelmän lisäksi kuvataan QDNAseq:n sovellutus translationaaliseen tutkimukseen ja kopiolukupoikkeamien tunnistamiseen alhaisen erilaistumisasteen glioomista. Todetaan kromosomin 10q häviämän yhteys huonoon ennusteeseen ja löydös vahvistetaan kahdessa riippumattomassa aineistossa. Tunnistettuja kopiolukupoikkeamia käytetään myös kasvaimien epäyhtenäisyyden ja ajallisen kehityksen tarkasteluun. Havaitaan kyseiselle syöpätyypille yleisen 1p/19q-häviämän olevan ainoa kopiolukupoikkeama, joka on johdonmukaisesti joko läsnä taikka puuttuu läpi sekä koko alkuperäisen syöpäkasvaimen että mahdollisten uusiutumien. Havainto sopii nykynäkemykseen kyseisen poikkeaman synnystä hyvin varhaisessa vaiheessa kyseisen syöpätyypin kehitystä. Lopuksi tarkastellaan kuvattujen bioinformaattisten ohjelmistojen kehitys- ja sovellutusprosesseista opittuja asioita. Ohjelmistokehitysalan vakiintuneiden käytänteiden parempi tuntemus olisi ollut hyödyllistä, ja yhdessä toteutusyksityiskohtien tarkemman harkinnan kanssa voinut auttaa tuottamaan tarkoituksensa paremmin täyttäviä sekä helpommin kehitettäviä ja ylläpidettäviä… Afwijkingen in het aantal chromosomen, of delen van chromosomen, zijn een van de mechanismen die aanleiding geven tot het proliferatieve gedrag van kankercellen. Deze chromosomale afwijkingen kunnen worden gemeten met genomische technieken met een hoge resolutie. Deze technieken genereren zeer grote hoeveelheden data, die op maat gemaakte bioinformatische oplossingen vereisen voor analyse en databeheer. De twee meest relevante genomische technieken met hoge resolutie zijn microarrays en ‘next generation sequencing’. Hoofdstuk 1 van dit proefschrift behandelt de literatuur van de data-analyse voor chromosomale afwijkingen gemeten met microarrays of ‘next generation sequencing’. Het introduceert relevante bioinformatische concepten, beschrijft het analytische proces van ruwe data tot identificatie van numerieke chromosoomafwijkingen in individuele tumoren en het bioinformatisch onderzoek gericht op de betekenis van die afwijkingen in grote series tumoren. Hoofdstuk 2 tot en met 4 beschrijven drie nieuwe bioinformatische implementaties ontwikkeld voor de analyse van deze chromosomale afwijkingen in kanker. CanGEM (Hoofdstuk 2) is een publiek toegankelijke database voor het opslaan van ruwe en verwerkte chromosoomaantallen het kankeronderzoek. De inhoud van de database kan worden doorzocht op basis van zowel klinische als experimentele gegevens met betrekking tot chromosoomaantallen. Klinische gegevens worden verzameld met behulp van gecontroleerde woordenlijsten. Chromosoomaantallen worden verzameld als ruwe microarray data en begin- en eindpositie van de afwijkingen worden steeds opnieuw automatisch bepaald. Om de integratie van de data, die gemeten worden met microarrays van verschillende makelij, verder te faciliteren, wordt het aantal chromosomen per gen afgeleid voor ieder van de ca. 19.000 tot 20.000 menselijke genen. CGHpower (Hoofdstuk 3) is een methode om te berekenen hoeveel tumormonsters statistisch nodig zijn om verschillen en overeenkomsten in chromosomale afwijkingen tussen twee groepen tumoren te kunnen vergelijken. Er wordt een schatting gemaakt van de complexiteit van de afwijkingen in een bepaald type kanker met behulp van een beperkt aantal monsters. Vervolgens wordt geschat hoeveel tumoren nodig zijn om aan de statistische eisen te voldoen. CGHpower kan in de planningsfase van een subsidieaanvraag worden gebruikt als rechtvaardiging van de voorgestelde aantallen naar een subsidiegever, of kan gebruikt worden om te controleren of er voldoende aantallen tumoren in een experiment werden opgenomen. CGHpower wordt geëvalueerd met behulp van experimentele en gesimuleerde datasets. QDNAseq (Hoofdstuk 4) is een methode die een voorbewerkingstap maakt van ‘next generation sequencing’ data naar chromosoomaantallen in het genoom van een tumor, waarbij wordt uitgegaan van sequencing met een diepte van slechts 10\% van het gehele genoom. QDNAseq corrigeert de waargenomen genoomwijde dekking voor systematische fouten en faciliteert de mogelijkheid om onregelmatige gebieden in het genoom te verwijderen. Een lijst van dergelijke systematische fouten en onregelmatige gebieden is afgeleid van publieke data die openbaar werd gemaakt door het “1000 Genomes Project”. QDNAseq wordt geëvalueerd ten opzichte van de microarraytechniek en andere gepubliceerde software voor de analyse van numerieke chromosoomafwijkingen met behulp van ‘next generation sequencing’. Tenslotte worden de uitkomsten van QDNAseq op ‘next generation sequencing’ data vergeleken met theoretische statistisch verwachte resultaten. In het voorlaatste hoofdstuk (Chapter 5) wordt QDNAseq toegepast op translationeel onderzoek dat tot doel heeft afwijkingen in het aantal chromosomen of delen daarvan te identificeren bij tumoren van patiënten met laag-gradige gliomen. Chromosomale afwijkingen geïdentificeerd middels ‘next generation sequencing’ en QDNAseq worden gebruikt om associaties te bepalen met de overleving van de patiënt, de intratumorale heterogeniteit van de tumoren en de evolutie over tijd van deze tumoren. Een verlies van het distale deel van chromosoom 10q wordt in dit onderzoek geassocieerd met een slechte prognose. Deze bevinding kon worden gevalideerd in twee onafhankelijke patiëntenseries. Uit de beoordeling van intratumorale heterogeniteit en tumorevolutie blijkt tenslotte dat verlies van chromosoom 1p samen met 19q de enige afwijking is die consistent aan- of afwezig is in de tumoren. Net als bij de drie beschreven implementaties voor de analyse van chromosomale afwijkingen in kanker, wordt veel bioinformatisch onderzoek uitgevoerd in academische groepen. De discussie (Hoofdstuk 6) behandelt de opgedane ervaringen met betrekking tot het ontwikkelingsproces en de toepassing van bioinformatische oplossingen.