Mats G. Gustafsson

Cross-validation and bootstrapping are unreliable in small sample classification

The interest in statistical classification for critical applications such as diagnoses of patient samples based on supervised learning is rapidly growing. To gain acceptance in applications where the subsequent decisions have serious consequences, e.g. choice of cancer therapy, any such decision support system must come with a reliable performance estimate. Tailored for small sample problems, cross-validation (CV) and bootstrapping (BTS) have been the most commonly used methods to determine such estimates in virtually all branches of science for the last 20 years. Here, we address the often overlooked fact that the uncertainty in a point estimate obtained with CV and BTS is unknown and quite large for small sample classification problems encountered in biomedical applications and elsewhere. To avoid this fundamental problem of employing CV and BTS, until improved alternatives have been established, we suggest that the final classification performance always should be reported in the form of a Bayesian confidence interval obtained from a simple holdout test or using some other method that yields conservative measures of the uncertainty.

Genome-wide signatures of differential DNA methylation in pediatric acute lymphoblastic leukemia

BackgroundAlthough aberrant DNA methylation has been observed previously in acute lymphoblastic leukemia (ALL), the patterns of differential methylation have not been comprehensively determined in all subtypes of ALL on a genome-wide scale. The relationship between DNA methylation, cytogenetic background, drug resistance and relapse in ALL is poorly understood.ResultsWe surveyed the DNA methylation levels of 435,941 CpG sites in samples from 764 children at diagnosis of ALL and from 27 children at relapse. This survey uncovered four characteristic methylation signatures. First, compared with control blood cells, the methylomes of ALL cells shared 9,406 predominantly hypermethylated CpG sites, independent of cytogenetic background. Second, each cytogenetic subtype of ALL displayed a unique set of hyper- and hypomethylated CpG sites. The CpG sites that constituted these two signatures differed in their functional genomic enrichment to regions with marks of active or repressed chromatin. Third, we identified subtype-specific differential methylation in promoter and enhancer regions that were strongly correlated with gene expression. Fourth, a set of 6,612 CpG sites was predominantly hypermethylated in ALL cells at relapse, compared with matched samples at diagnosis. Analysis of relapse-free survival identified CpG sites with subtype-specific differential methylation that divided the patients into different risk groups, depending on their methylation status.ConclusionsOur results suggest an important biological role for DNA methylation in the differences between ALL subtypes and in their clinical outcome after treatment.

DNA methylation for subtype classification and prediction of treatment outcome in patients with childhood acute lymphoblastic leukemia

Despite improvements in the prognosis of childhood acute lymphoblastic leukemia (ALL), subgroups of patients would benefit from alternative treatment approaches. Our aim was to identify genes with DNA methylation profiles that could identify such groups. We determined the methylation levels of 1320 CpG sites in regulatory regions of 416 genes in cells from 401 children diagnosed with ALL. Hierarchical clustering of 300 CpG sites distinguished between T-lineage ALL and B-cell precursor (BCP) ALL and between the main cytogenetic subtypes of BCP ALL. It also stratified patients with high hyperdiploidy and t(12;21) ALL into 2 subgroups with different probability of relapse. By using supervised learning, we constructed multivariate classifiers by external cross-validation procedures. We identified 40 genes that consistently contributed to accurate discrimination between the main subtypes of BCP ALL and gene sets that discriminated between subtypes of ALL and between ALL and controls in pairwise classification analyses. We also identified 20 individual genes with DNA methylation levels that predicted relapse of leukemia. Thus, methylation analysis should be explored as a method to improve stratification of ALL patients. The genes highlighted in our study are not enriched to specific pathways, but the gene expression levels are inversely correlated to the methylation levels.

Identification of molecular mechanisms for cellular drug resistance by combining drug activity and gene expression profiles

Acquired drug resistance is a major problem in cancer treatment. To explore the genes involved in chemosensitivity and resistance, 10 human tumour cell lines, including parental cells and resistant subtypes selected for resistance against doxorubicin, melphalan, teniposide and vincristine, were profiled for mRNA expression of 7400 genes using cDNA microarray technology. The drug activity of 66 cancer agents was evaluated on the cell lines, and correlations between drug activity and gene expression were calculated and ranked. Hierarchical clustering of drugs based on their drug–gene correlations yielded clusters of drugs with similar mechanism of action. Genes correlated with drug sensitivity and resistance were imported into the PathwayAssist software to identify putative molecular pathways involved. A substantial number of both proapoptotic and antiapoptotic genes such as signal transducer and activator of transcription 1, mitogen-activated protein kinase 1 and focal adhesion kinase were found to be associated to drug resistance, whereas genes linked to cell cycle control and proliferation, such as cell division cycle 25A and signal transducer of activator of transcription 5A, were associated to general drug sensitivity. The results indicate that combined information from drug activity and gene expression in a resistance-based cell line panel may provide new knowledge of the genes involved in anticancer drug resistance and become a useful tool in drug development.

ProteinSeq: High-Performance Proteomic Analyses by Proximity Ligation and Next Generation Sequencing

Despite intense interest, methods that provide enhanced sensitivity and specificity in parallel measurements of candidate protein biomarkers in numerous samples have been lacking. We present herein a multiplex proximity ligation assay with readout via realtime PCR or DNA sequencing (ProteinSeq). We demonstrate improved sensitivity over conventional sandwich assays for simultaneous analysis of sets of 35 proteins in 5 µl of blood plasma. Importantly, we observe a minimal tendency to increased background with multiplexing, compared to a sandwich assay, suggesting that higher levels of multiplexing are possible. We used ProteinSeq to analyze proteins in plasma samples from cardiovascular disease (CVD) patient cohorts and matched controls. Three proteins, namely P-selectin, Cystatin-B and Kallikrein-6, were identified as putative diagnostic biomarkers for CVD. The latter two have not been previously reported in the literature and their potential roles must be validated in larger patient cohorts. We conclude that ProteinSeq is promising for screening large numbers of proteins and samples while the technology can provide a much-needed platform for validation of diagnostic markers in biobank samples and in clinical use.

Phylogeny of theAsterales sensu lato based onrbcL sequences with particular reference to theGoodeniaceae

TherbcL gene of 25 taxa was sequenced and analyzed cladistically in order to define more precisely the orderAsterales s.l. and to reconstruct the phylogeny ofGoodeniaceae. The cladistic analyses show that theAsterales comprise the familiesAbrophyllaceae, Alseuosmiaceae, Argophyllaceae, Asteraceae, Calyceraceae, Campanulaceae s.l.,Donatiaceae, Goodeniaceae (includingBrunoniaceae),Menyanthaceae, Pentaphragmataceae, andStylidiaceae. Abrophyllaceae, Alseuosmiaceae, Brunoniaceae, andDonatiaceae have previously not been studied in this respect. Within theGoodeniaceae, four groups supported by therbcL data can be distinguished: the genusLechenaultia, theAnthotium-Dampiera-group, the genusBrunonia, and a group formed by the remaining genera, theScaevola-Goodenia-group.

Supervised identification of allergen-representative peptides for in silico detection of potentially allergenic proteins

MOTIVATION Identification of potentially allergenic proteins is needed for the safety assessment of genetically modified foods, certain pharmaceuticals and various other products on the consumer market. Current methods in bioinformatic allergology exploit common features among allergens for the detection of amino acid sequences of potentially allergenic proteins. Features for identification still unexplored include the motifs occurring commonly in allergens, but rarely in ordinary proteins. In this paper, we present an algorithm for the identification of such motifs with the purpose of biocomputational detection of amino acid sequences of potential allergens. RESULTS Identification of allergen-representative peptides (ARPs) with low or no occurrence in proteins lacking allergenic properties is the essential component of our new method, designated DASARP (Detection based on Automated Selection of Allergen-Representative Peptide). This approach consistently outperforms the criterion based on identical peptide match for predicting allergenicity recommended by ILSI/IFBC and FAO/WHO and shows results comparable to the alignment-based criterion as outlined by FAO/WHO. AVAILABILITY The detection software and the ARP set needed for the analysis of a query protein reported here are properties of the Swedish National Food Agency and are available upon request. The protein sequence sets used in this work are publicly available on http://www.slv.se/templatesSLV/SLV_Page____9343.asp. Allergenicity assessment for specific protein sequences of interest is also possible via ulfh@slv.se

Multiplexed Homogeneous Proximity Ligation Assays for High-throughput Protein Biomarker Research in Serological Material

A high throughput protein biomarker discovery tool has been developed based on multiplexed proximity ligation assays in a homogeneous format in the sense of no washing steps. The platform consists of four 24-plex panels profiling 74 putative biomarkers with sub-pm sensitivity each consuming only 1 μl of human plasma sample. The system uses either matched monoclonal antibody pairs or the more readily available single batches of affinity purified polyclonal antibodies to generate the target specific reagents by covalently linking with unique nucleic acid sequences. These paired sequences are united by DNA ligation upon simultaneous target binding forming a PCR amplicon. Multiplex proximity ligation assays thereby converts multiple target analytes into real-time PCR amplicons that are individually quantified using microfluidic high capacity qPCR in nano liter volumes. The assay shows excellent specificity, even in multiplex, by its dual recognition feature, its proximity requirement, and most importantly by using unique sequence specific reporter fragments on both antibody-based probes. To illustrate the potential of this protein detection technology, a pilot biomarker research project was performed using biobanked plasma samples for the detection of colorectal cancer using a multivariate signature.

Molecular Markers for Discrimination of Benign and Malignant Follicular Thyroid Tumors

Objective: To identify molecular markers useful for the diagnostic discrimination of benign and malignant follicular thyroid tumors. Methods: A panel of thyroid tumors was characterized with expression profiling using cDNA microarrays. A robust algorithm for gene selection was developed to identify molecular markers useful for the classification of heterogeneous tumor classes. The study included tumor tissue specimens from 10 patients with benign follicular adenomas and from 10 with malignant tumors. The malignant tumors mainly consisted of clinically relevant minimally invasive follicular carcinomas. The mRNA expression level of a candidate gene, FHL1, was evaluated in an independent series of 61 tumors. Results: 22 gene expression markers were identified as differentially expressed. Several of the identified genes, for example DIO1, CITED1, CA12 and FN1, have previously been observed as differentially expressed in various thyroid tumors. FHL1 was significantly underexpressed in carcinomas compared to adenomas in the independent panel of tumors. The results indicate that a small number of genes can be useful to distinguish follicular adenomas from follicular carcinomas. Conclusions: Our findings clearly corroborate previous studies and identify novel candidate molecular markers. These genes have the potential for molecular classification of follicular thyroid tumors and for providing improved understanding of the molecular mechanisms involved in thyroid malignancies.

Independent component analysis yields chemically interpretable latent variables in multivariate regression

This work shows that independent component analysis (ICA) can be used to obtain statistically independent and, therefore, chemically interpretable latent variables (LVs) in multivariate regression. Two novel algorithms based on ICA are introduced and compared with two classical methods on simulated data: principal component regression and partial least-squares regression. All methods compared yield accurate predictions, but only those based on ICA yield LVs that are chemically interpretable. Practical limitations of ICA-based regression with respect to the underlying assumptions, sample size, and measurement noise are discussed and illustrated by means of simulations.