Artificial neural network models to predict nodal status in clinically node-negative breast cancer.

BACKGROUND: Sentinel lymph node biopsy (SLNB) is standard staging procedure for nodal status in breast cancer, but lacks therapeutic benefit for patients with benign sentinel nodes. For patients with positive sentinel nodes, individualized surgical strategies are applied depending on the extent of nodal involvement. Preoperative prediction of nodal status is thus important for individualizing axillary surgery avoiding unnecessary surgery. We aimed to predict nodal status in clinically node-negative breast cancer and identify candidates for SLNB omission by including patient-related and pathological characteristics into artificial neural network (ANN) models. METHODS: Patients with primary breast cancer were consecutively included between January 1, 2009 and December 31, 2012 in a prospectively maintained pathology database. Clinical- and radiological data were extracted from patient's files and only clinically node-negative patients constituted the final study cohort. ANN-based models for nodal prediction were constructed including 15 risk variables for nodal status. Area under the receiver operating characteristic curve (AUC) and Hosmer-Lemeshow goodness-of-fit test (HL) were used to assess performance and calibration of three predictive ANN-based models for no lymph node metastasis (N0), metastases in 1-3 lymph nodes (N1) and metastases in ≥ 4 lymph nodes (N2). Linear regression models for nodal prediction were calculated for comparison. RESULTS: Eight hundred patients (N0, n = 514; N1, n = 232; N2, n = 54) were included. Internally validated AUCs for N0 versus N+ was 0.740 (95% CI = 0.723-0.758); median HL was 9.869 (P = 0.274), for N1 versus N0, 0.705 (95% CI = 0.686-0.724; median HL: 7.421; P = 0.492) and for N2 versus N0 and N1, 0.747 (95% CI = 0.728-0.765; median HL: 9.220; P = 0.324). Tumor size and vascular invasion were top-ranked predictors of all three end-points, followed by estrogen receptor status and lobular cancer for prediction of N2. For each end-point, ANN models showed better discriminatory performance than multivariable logistic regression models. Accepting a false negative rate (FNR) of 10% for predicting N0 by the ANN model, SLNB could have been abstained in 27.25% of patients with clinically node-negative axilla. CONCLUSIONS: In this retrospective study, ANN showed promising result as decision-supporting tools for estimating nodal disease. If prospectively validated, patients least likely to have nodal metastasis could be spared SLNB using predictive models. TRIAL REGISTRATION: Registered in the ISRCTN registry with study ID ISRCTN14341750 . Date of registration 23/11/2018. Retrospectively registered.

Identification and validation of single-sample breast cancer radiosensitivity gene expression predictors.

BACKGROUND: Adjuvant radiotherapy is the standard of care after breast-conserving surgery for primary breast cancer, despite a majority of patients being over- or under-treated. In contrast to adjuvant endocrine therapy and chemotherapy, no diagnostic tests are in clinical use that can stratify patients for adjuvant radiotherapy. This study presents the development and validation of a targeted gene expression assay to predict the risk of ipsilateral breast tumor recurrence and response to adjuvant radiotherapy after breast-conserving surgery in primary breast cancer. METHODS: Fresh-frozen primary tumors from 336 patients radically (clear margins) operated on with breast-conserving surgery with or without radiotherapy were collected. Patients were split into a discovery cohort (N = 172) and a validation cohort (N = 164). Genes predicting ipsilateral breast tumor recurrence in an Illumina HT12 v4 whole transcriptome analysis were combined with genes identified in the literature (248 genes in total) to develop a targeted radiosensitivity assay on the Nanostring nCounter platform. Single-sample predictors for ipsilateral breast tumor recurrence based on a k-top scoring pairs algorithm were trained, stratified for estrogen receptor (ER) status and radiotherapy. Two previously published profiles, the radiosensitivity signature of Speers et al., and the 10-gene signature of Eschrich et al., were also included in the targeted panel. RESULTS: Derived single-sample predictors were prognostic for ipsilateral breast tumor recurrence in radiotherapy-treated ER+ patients (AUC 0.67, p = 0.01), ER+ patients without radiotherapy (AUC = 0.89, p = 0.02), and radiotherapy-treated ER- patients (AUC = 0.78, p < 0.001). Among ER+ patients, radiotherapy had an excellent effect on tumors classified as radiosensitive (p < 0.001), while radiotherapy had no effect on tumors classified as radioresistant (p = 0.36) and there was a high risk of ipsilateral breast tumor recurrence (55% at 10 years). Our single-sample predictors developed in ER+ tumors and the radiosensitivity signature correlated with proliferation, while single-sample predictors developed in ER- tumors correlated with immune response. The 10-gene signature negatively correlated with both proliferation and immune response. CONCLUSIONS: Our targeted single-sample predictors were prognostic for ipsilateral breast tumor recurrence and have the potential to stratify patients for adjuvant radiotherapy. The correlation of models with biology may explain the different performance in subgroups of breast cancer.

A multicenter study investigating the molecular fingerprint of psychological resilience in breast cancer patients: study protocol of the SCAN-B resilience study.

BACKGROUND: Individual patients differ in their psychological response when receiving a cancer diagnosis, in this case breast cancer. Given the same disease burden, some patients master the situation well, while others experience a great deal of stress, depression and lowered quality of life. Patients with high psychological resilience are likely to experience fewer stress reactions and better adapt to and manage the life threat and the demanding treatment that follows the diagnosis. If this phenomenon of mastering difficult situations is reflected also in biomolecular processes is not much studied, nor has its capacity for impacting the cancer prognosis been addressed. This project specifically aims, for the first time, to investigate how a breast cancer patient's psychological resilience is coupled to biomolecular parameters using advanced "omics" and, as a secondary aim, whether it relates to prognosis and quality of life one year after diagnosis. METHOD: The study population consists of newly diagnosed breast cancer patients enrolled in the Sweden Cancerome Analysis Network - Breast (SCAN-B) at four hospitals in Sweden. At the time of cancer diagnosis, the patient fills out the standardized method to measure psychological resilience, the "Connor-Davidson Resilience scale" (CD-RISC), the quality of life measure SF-36, as well as providing social and socioeconomic variables. In addition, one blood sample is collected. At the one-year follow-up, the patient will be subjected to the same assessments, and we also collect information regarding smoking, exercise habits, and BMI, as well as patients' trust in the treatment and their satisfaction with the care and treatment. DISCUSSION: This explorative hypothesis-generating project will pave the way for larger validation studies, potentially leading to a standardized method of measuring psychological resilience as an important parameter in cancer care. Revealing the body-mind interaction, in terms of psychological resilience and quality of life, will herald the development of truly personalized psychosocial care and cancer intervention treatment strategies. TRIAL REGISTRATION: This is a retrospectively registered trial at ClinicalTrials.gov, ID: NCT03430492 on February 6, 2018.

SOX11 and TP53 add prognostic information to MIPI in a homogenously treated cohort of mantle cell lymphoma--a Nordic Lymphoma Group study.

Mantle cell lymphoma (MCL) is an aggressive B cell lymphoma, where survival has been remarkably improved by use of protocols including high dose cytarabine, rituximab and autologous stem cell transplantation, such as the Nordic MCL2/3 protocols. In 2008, a MCL international prognostic index (MIPI) was created to enable stratification of the clinical diverse MCL patients into three risk groups. So far, use of the MIPI in clinical routine has been limited, as it has been shown that it inadequately separates low and intermediate risk group patients. To improve outcome and minimize treatment-related morbidity, additional parameters need to be evaluated to enable risk-adapted treatment selection. We have investigated the individual prognostic role of the MIPI and molecular markers including SOX11, TP53 (p53), MKI67 (Ki-67) and CCND1 (cyclin D1). Furthermore, we explored the possibility of creating an improved prognostic tool by combining the MIPI with information on molecular markers. SOX11 was shown to significantly add prognostic information to the MIPI, but in multivariate analysis TP53 was the only significant independent molecular marker. Based on these findings, we propose that TP53 and SOX11 should routinely be assessed and that a combined TP53/MIPI score may be used to guide treatment decisions.

Transcriptional regulation of lineage commitment--a stochastic model of cell fate decisions.

Molecular mechanisms employed by individual multipotent cells at the point of lineage commitment remain largely uncharacterized. Current paradigms span from instructive to noise-driven mechanisms. Of considerable interest is also whether commitment involves a limited set of genes or the entire transcriptional program, and to what extent gene expression configures multiple trajectories into commitment. Importantly, the transient nature of the commitment transition confounds the experimental capture of committing cells. We develop a computational framework that simulates stochastic commitment events, and affords mechanistic exploration of the fate transition. We use a combined modeling approach guided by gene expression classifier methods that infers a time-series of stochastic commitment events from experimental growth characteristics and gene expression profiling of individual hematopoietic cells captured immediately before and after commitment. We define putative regulators of commitment and probabilistic rules of transition through machine learning methods, and employ clustering and correlation analyses to interrogate gene regulatory interactions in multipotent cells. Against this background, we develop a Monte Carlo time-series stochastic model of transcription where the parameters governing promoter status, mRNA production and mRNA decay in multipotent cells are fitted to experimental static gene expression distributions. Monte Carlo time is converted to physical time using cell culture kinetic data. Probability of commitment in time is a function of gene expression as defined by a logistic regression model obtained from experimental single-cell expression data. Our approach should be applicable to similar differentiating systems where single cell data is available. Within our system, we identify robust model solutions for the multipotent population within physiologically reasonable values and explore model predictions with regard to molecular scenarios of entry into commitment. The model suggests distinct dependencies of different commitment-associated genes on mRNA dynamics and promoter activity, which globally influence the probability of lineage commitment.

Facilitating clinically relevant skin tumor diagnostics with spectroscopy-driven machine learning.

In the dawning era of artificial intelligence (AI), health care stands to undergo a significant transformation with the increasing digitalization of patient data. Digital imaging, in particular, will serve as an important platform for AI to aid decision making and diagnostics. A growing number of studies demonstrate the potential of automatic pre-surgical skin tumor delineation, which could have tremendous impact on clinical practice. However, current methods rely on having ground truth images in which tumor borders are already identified, which is not clinically possible. We report a novel approach where hyperspectral images provide spectra from small regions representing healthy tissue and tumor, which are used to generate prediction maps using artificial neural networks (ANNs), after which a segmentation algorithm automatically identifies the tumor borders. This circumvents the need for ground truth images, since an ANN model is trained with data from each individual patient, representing a more clinically relevant approach.

A community effort to identify and correct mislabeled samples in proteogenomic studies.

Sample mislabeling or misannotation has been a long-standing problem in scientific research, particularly prevalent in large-scale, multi-omic studies due to the complexity of multi-omic workflows. There exists an urgent need for implementing quality controls to automatically screen for and correct sample mislabels or misannotations in multi-omic studies. Here, we describe a crowdsourced precisionFDA NCI-CPTAC Multi-omics Enabled Sample Mislabeling Correction Challenge, which provides a framework for systematic benchmarking and evaluation of mislabel identification and correction methods for integrative proteogenomic studies. The challenge received a large number of submissions from domestic and international data scientists, with highly variable performance observed across the submitted methods. Post-challenge collaboration between the top-performing teams and the challenge organizers has created an open-source software, COSMO, with demonstrated high accuracy and robustness in mislabeling identification and correction in simulated and real multi-omic datasets.

Identification of uniquely expressed transcription factors in highly purified B-cell lymphoma samples.

Transcription factors (TFs) are critical for B-cell differentiation, affecting gene expression both by repression and transcriptional activation. Still, this information is not used for classification of B-cell lymphomas (BCLs). Traditionally, BCLs are diagnosed based on a phenotypic resemblance to normal B-cells; assessed by immunohistochemistry or flow cytometry, by using a handful of phenotypic markers. In the last decade, diagnostic and prognostic evaluation has been facilitated by global gene expression profiling (GEP), providing a new powerful means for the classification, prediction of survival, and response to treatment of lymphomas. However, most GEP studies have typically been performed on whole tissue samples, containing varying degrees of tumor cell content, which results in uncertainties in data analysis. In this study, global GEP analyses were performed on highly purified, flow-cytometry sorted tumor-cells from eight subgroups of BCLs. This enabled identification of TFs that can be uniquely associated to the tumor cells of chronic lymphocytic leukemia (CLL), diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), hairy cell leukemia (HCL), and mantle cell lymphoma (MCL). The identified transcription factors influence both the global and specific gene expression of the BCLs and have possible implications for diagnosis and treatment.

Accounting for one-channel depletion improves missing value imputation in 2-dye microarray data.

BACKGROUND: For 2-dye microarray platforms, some missing values may arise from an un-measurably low RNA expression in one channel only. Information of such "one-channel depletion" is so far not included in algorithms for imputation of missing values. RESULTS: Calculating the mean deviation between imputed values and duplicate controls in five datasets, we show that KNN-based imputation gives a systematic bias of the imputed expression values of one-channel depleted spots. Evaluating the correction of this bias by cross-validation showed that the mean square deviation between imputed values and duplicates were reduced up to 51%, depending on dataset. CONCLUSION: By including more information in the imputation step, we more accurately estimate missing expression values.

Genes associated with the tumour microenvironment are differentially expressed in cured versus primary chemotherapy-refractory diffuse large B-cell lymphoma.

In order to identify genes associated with primary chemotherapy-resistance, gene expression profiles (GEP) in tumour tissue from 37 patients with de novo diffuse large B-cell lymphoma (DLBCL), stage II-IV, either in continuous complete remission (n = 24) or with progressive disease during primary treatment (n = 13), were examined using spotted 55K oligonucleotide arrays. Immunohistochemistry was used for confirmation at the protein level. The top 86 genes that best discriminated between the two cohorts were chosen for further analysis. Only seven of 86 genes were overexpressed in the refractory cohort, e.g. RABGGTB and POLE, both potential targets for drug intervention. Seventy-nine of 86 genes were overexpressed in the cured cohort and mainly coded for proteins expressed in the tumour microenvironment, many of them involved in proteolytic activity and remodelling of extra cellular matrix. Furthermore, major histocompatibility complex class I molecules, CD3D and ICAM1 were overexpressed, indicating an enhanced immunological reaction. Immunohistochemistry confirmed the GEP results. The frequency of tumour infiltrating lymphocytes, macrophages, and reactive cells expressing ICAM-1, lysozyme, cathepsin D, urokinase plasminogen activator receptor, signal transducer and activator of transcription 1, and galectin-3 was higher in the cured cohort. These findings indicate that a reactive microenvironment has an impact on the outcome of chemotherapy in DLBCL.

Estrogen receptor beta expression is associated with tamoxifen response in ERalpha-negative breast carcinoma.

PURPOSE: Endocrine therapies, such as tamoxifen, are commonly given to most patients with estrogen receptor (ERalpha)-positive breast carcinoma but are not indicated for persons with ERalpha-negative cancer. The factors responsible for response to tamoxifen in 5% to 10% of patients with ERalpha-negative tumors are not clear. The aim of the present study was to elucidate the biology and prognostic role of the second ER, ERbeta, in patients treated with adjuvant tamoxifen. EXPERIMENTAL DESIGN: We investigated ERbeta by immunohistochemistry in 353 stage II primary breast tumors from patients treated with 2 years adjuvant tamoxifen, and generated gene expression profiles for a representative subset of 88 tumors. RESULTS: ERbeta was associated with increased survival (distant disease-free survival, P = 0.01; overall survival, P = 0.22), and in particular within ERalpha-negative patients (P = 0.003; P = 0.04), but not in the ERalpha-positive subgroup (P = 0.49; P = 0.88). Lack of ERbeta conferred early relapse (hazard ratio, 14; 95% confidence interval, 1.8-106; P = 0.01) within the ERalpha-negative subgroup even after adjustment for other markers. ERalpha was an independent marker only within the ERbeta-negative tumors (hazard ratio, 0.44; 95% confidence interval, 0.21-0.89; P = 0.02). An ERbeta gene expression profile was identified and was markedly different from the ERalpha signature. CONCLUSION: Expression of ERbeta is an independent marker for favorable prognosis after adjuvant tamoxifen treatment in ERalpha-negative breast cancer patients and involves a gene expression program distinct from ERalpha. These results may be highly clinically significant, because in the United States alone, approximately 10,000 women are diagnosed annually with ERalpha-negative/ERbeta-positive breast carcinoma and may benefit from adjuvant tamoxifen.

Diagnostic and prognostic gene expression signatures in 177 soft tissue sarcomas: hypoxia-induced transcription profile signifies metastatic potential.

BACKGROUND: Soft tissue sarcoma (STS) diagnosis is challenging because of a multitude of histopathological subtypes, different genetic characteristics, and frequent intratumoral pleomorphism. One-third of STS metastasize and current risk-stratification is suboptimal, therefore, novel diagnostic and prognostic markers would be clinically valuable. We assessed the diagnostic and prognostic value of array-based gene expression profiles using 27 k cDNA microarrays in 177, mainly high-grade, STS of 13 histopathological subtypes. RESULTS: Unsupervised analysis resulted in two major clusters--one mainly containing STS characterized by type-specific genetic alterations and the other with a predominance of genetically complex and pleomorphic STS. Synovial sarcomas, myxoid/round-cell liposarcomas, and gastrointestinal stromal tumors clustered tightly within the former cluster and discriminatory signatures for these were characterized by developmental genes from the EGFR, FGFR, Wnt, Notch, Hedgehog, RAR and KIT signaling pathways. The more pleomorphic STS subtypes, e.g. leiomyosarcoma, malignant fibrous histiocytoma/undifferentiated pleomorphic sarcoma and dedifferentiated/pleomorphic liposarcoma, were part of the latter cluster and were characterized by relatively heterogeneous profiles, although subclusters herein were identified. A prognostic signature partly characterized by hypoxia-related genes was identified among 89 genetically complex pleomorphic primary STS and could, in a multivariate analysis including established prognostic markers, independently predict the risk of metastasis with a hazard ratio of 2.2 (P = 0.04). CONCLUSION: Diagnostic gene expression profiles linking signaling pathways to the different STS subtypes were demonstrated and a hypoxia-induced metastatic profile was identified in the pleomorphic, high-grade STS. These findings verify diagnostic utility and application of expression data for improved selection of high-risk STS patients.

Gene expression profilers and conventional clinical markers to predict distant recurrences for premenopausal breast cancer patients after adjuvant chemotherapy.

A large proportion of breast cancer patients are treated with adjuvant chemotherapy after the primary operation, but some will recur in spite of this treatment. In order to achieve an improved and more individualised therapy, our knowledge in mechanisms for drug resistance needs to be increased. We have investigated to what extent cDNA microarray measurements could distinguish the likelihood of recurrences after adjuvant CMF (cyclophosphamide, methotrexate and 5-fluorouracil) treatment of premenopausal, lymph node positive breast cancer patients, and have also compared this with the corresponding performance when using conventional clinical variables. We tried several gene selection strategies, and built classifiers using the resulting gene lists. The best performing classifier with odds ratio (OR)=6.5 (95% confidence interval (CI)=1.4-62) did not outperform corresponding classifiers based on clinical variables. For the clinical variables, calibrated on the samples, either using all the clinical parameters or the Nottingham Prognostic Index (NPI) parameters, the areas under the receiver operating characteristics (ROC) curve were 0.78 and 0.79, respectively. The ORs at 90% sensitivity were 15 (95% CI=3.1-140) and 10 (95% CI=2.1-97), respectively. Our data have provided evidence for a comparable prediction of clinical outcome in CMF-treated breast cancer patients using conventional clinical variables and gene expression based markers.

Pulsatile blood flow, shear force, energy dissipation and Murray's Law.

BACKGROUND: Murray's Law states that, when a parent blood vessel branches into daughter vessels, the cube of the radius of the parent vessel is equal to the sum of the cubes of the radii of daughter blood vessels. Murray derived this law by defining a cost function that is the sum of the energy cost of the blood in a vessel and the energy cost of pumping blood through the vessel. The cost is minimized when vessel radii are consistent with Murray's Law. This law has also been derived from the hypothesis that the shear force of moving blood on the inner walls of vessels is constant throughout the vascular system. However, this derivation, like Murray's earlier derivation, is based on the assumption of constant blood flow. METHODS: To determine the implications of the constant shear force hypothesis and to extend Murray's energy cost minimization to the pulsatile arterial system, a model of pulsatile flow in an elastic tube is analyzed. A new and exact solution for flow velocity, blood flow rate and shear force is derived. RESULTS: For medium and small arteries with pulsatile flow, Murray's energy minimization leads to Murray's Law. Furthermore, the hypothesis that the maximum shear force during the cycle of pulsatile flow is constant throughout the arterial system implies that Murray's Law is approximately true. The approximation is good for all but the largest vessels (aorta and its major branches) of the arterial system. CONCLUSION: A cellular mechanism that senses shear force at the inner wall of a blood vessel and triggers remodeling that increases the circumference of the wall when a shear force threshold is exceeded would result in the observed scaling of vessel radii described by Murray's Law.

Predicting continuous values of prognostic markers in breast cancer from microarray gene expression profiles.

The prognostic and treatment-predictive markers currently in use for breast cancer are commonly based on the protein levels of individual genes (e.g., steroid receptors) or aspects of the tumor phenotype, such as histological grade and percentage of cells in the DNA synthesis phase of the cell cycle. Microarrays have previously been used to classify binary classes in breast cancer such as estrogen receptor (ER)-alpha status. To test whether the properties and specific values of conventional prognostic markers are encoded within tumor gene expression profiles, we have analyzed 48 well-characterized primary tumors from lymph node-negative breast cancer patients using 6728-element cDNA microarrays. In the present study, we used artificial neural networks trained with tumor gene expression data to predict the ER protein values on a continuous scale. Furthermore, we determined a gene expression profile-directed threshold for ER protein level to redefine the cutoff between ER-positive and ER-negative classes that may be more biologically relevant. With a similar approach, we studied the prediction of other prognostic parameters such as percentage cells in the S phase of the cell cycle (SPF), histological grade, DNA ploidy status, and progesterone receptor status. Interestingly, there was a consistent reciprocal relationship in expression levels of the genes important for both ER and SPF prediction. This and similar studies may be used to increase our understanding of the biology underlying these markers as well as to improve the currently available prognostic markers for breast cancer.

Finding Risk Groups by Optimizing Artificial Neural Networks on the Area under the Survival Curve Using Genetic Algorithms.

We investigate a new method to place patients into risk groups in censored survival data. Properties such as median survival time, and end survival rate, are implicitly improved by optimizing the area under the survival curve. Artificial neural networks (ANN) are trained to either maximize or minimize this area using a genetic algorithm, and combined into an ensemble to predict one of low, intermediate, or high risk groups. Estimated patient risk can influence treatment choices, and is important for study stratification. A common approach is to sort the patients according to a prognostic index and then group them along the quartile limits. The Cox proportional hazards model (Cox) is one example of this approach. Another method of doing risk grouping is recursive partitioning (Rpart), which constructs a decision tree where each branch point maximizes the statistical separation between the groups. ANN, Cox, and Rpart are compared on five publicly available data sets with varying properties. Cross-validation, as well as separate test sets, are used to validate the models. Results on the test sets show comparable performance, except for the smallest data set where Rpart's predicted risk groups turn out to be inverted, an example of crossing survival curves. Cross-validation shows that all three models exhibit crossing of some survival curves on this small data set but that the ANN model manages the best separation of groups in terms of median survival time before such crossings. The conclusion is that optimizing the area under the survival curve is a viable approach to identify risk groups. Training ANNs to optimize this area combines two key strengths from both prognostic indices and Rpart. First, a desired minimum group size can be specified, as for a prognostic index. Second, the ability to utilize non-linear effects among the covariates, which Rpart is also able to do.

Single-Cell Network Analysis Identifies DDIT3 as a Nodal Lineage Regulator in Hematopoiesis.

We explore cell heterogeneity during spontaneous and transcription-factor-driven commitment for network inference in hematopoiesis. Since individual genes display discrete OFF states or a distribution of ON levels, we compute and combine pairwise gene associations from binary and continuous components of gene expression in single cells. Ddit3 emerges as a regulatory node with positive linkage to erythroid regulators and negative association with myeloid determinants. Ddit3 loss impairs erythroid colony output from multipotent cells, while forcing Ddit3 in granulo-monocytic progenitors (GMPs) enhances self-renewal and impedes differentiation. Network analysis of Ddit3-transduced GMPs reveals uncoupling of myeloid networks and strengthening of erythroid linkages. RNA sequencing suggests that Ddit3 acts through development or stabilization of a precursor upstream of GMPs with inherent Meg-E potential. The enrichment of Gata2 target genes in Ddit3-dependent transcriptional responses suggests that Ddit3 functions in an erythroid transcriptional network nucleated by Gata2.

Comparing functional annotation analyses with Catmap.

BACKGROUND: Ranked gene lists from microarray experiments are usually analysed by assigning significance to predefined gene categories, e.g., based on functional annotations. Tools performing such analyses are often restricted to a category score based on a cutoff in the ranked list and a significance calculation based on random gene permutations as null hypothesis. RESULTS: We analysed three publicly available data sets, in each of which samples were divided in two classes and genes ranked according to their correlation to class labels. We developed a program, Catmap (available for download at http://bioinfo.thep.lu.se/Catmap), to compare different scores and null hypotheses in gene category analysis, using Gene Ontology annotations for category definition. When a cutoff-based score was used, results depended strongly on the choice of cutoff, introducing an arbitrariness in the analysis. Comparing results using random gene permutations and random sample permutations, respectively, we found that the assigned significance of a category depended strongly on the choice of null hypothesis. Compared to sample label permutations, gene permutations gave much smaller p-values for large categories with many coexpressed genes. CONCLUSIONS: In gene category analyses of ranked gene lists, a cutoff independent score is preferable. The choice of null hypothesis is very important; random gene permutations does not work well as an approximation to sample label permutations.

Expression profiling to predict outcome in breast cancer: the influence of sample selection.

Gene expression profiling of tumors using DNA microarrays is a promising method for predicting prognosis and treatment response in cancer patients. It was recently reported that expression profiles of sporadic breast cancers could be used to predict disease recurrence better than currently available clinical and histopathological prognostic factors. Having observed an overlap in those data between the genes that predict outcome and those that predict estrogen receptor-alpha status, we examined their predictive power in an independent data set. We conclude that it may be important to define prognostic expression profiles separately for estrogen receptor-alpha-positive and estrogen receptor-alpha-negative tumors.

"Good Old" clinical markers have similar power in breast cancer prognosis as microarray gene expression profilers.

We compared the power of gene expression measurements with that of conventional prognostic markers, i.e., clinical, histopathological, and cell biological parameters, for predicting distant metastases in breast cancer patients using both established prognostic indices (e.g., the Nottingham Prognostic Index (NPI)) and novel combinations of conventional markers. We used publicly available data on 97 patients, and the performance of metastasis prediction was represented by receiver operating characteristic (ROC) areas and Kaplan-Meier plots. The gene expression profiler did not perform noticeably better than indices constructed from the clinical variables, e.g., the well established NPI. When analysing separately subgroups, according to the oestrogen receptor (ER) status both approaches could predict clinical outcome more easily for the ER-positive than for the ER-negative cohort. Given the time it may take before microarray processing is used worldwide, particularly due to the costs and the lack of standards, it is important to pursue research using conventional markers. Our analysis suggests that it might be possible to improve the combination of different conventional prognostic markers into one prognostic index.