Heuristic-enabled active machine learning: A case study of predicting essential developmental stage and immune response genes in Drosophila melanogaster.

Computational prediction of absolute essential genes using machine learning has gained wide attention in recent years. However, essential genes are mostly conditional and not absolute. Experimental techniques provide a reliable approach of identifying conditionally essential genes; however, experimental methods are laborious, time and resource consuming, hence computational techniques have been used to complement the experimental methods. Computational techniques such as supervised machine learning, or flux balance analysis are grossly limited due to the unavailability of required data for training the model or simulating the conditions for gene essentiality. This study developed a heuristic-enabled active machine learning method based on a light gradient boosting model to predict essential immune response and embryonic developmental genes in Drosophila melanogaster. We proposed a new sampling selection technique and introduced a heuristic function which replaces the human component in traditional active learning models. The heuristic function dynamically selects the unlabelled samples to improve the performance of the classifier in the next iteration. Testing the proposed model with four benchmark datasets, the proposed model showed superior performance when compared to traditional active learning models (random sampling and uncertainty sampling). Applying the model to identify conditionally essential genes, four novel essential immune response genes and a list of 48 novel genes that are essential in embryonic developmental condition were identified. We performed functional enrichment analysis of the predicted genes to elucidate their biological processes and the result evidence our predictions. Immune response and embryonic development related processes were significantly enriched in the essential immune response and embryonic developmental genes, respectively. Finally, we propose the predicted essential genes for future experimental studies and use of the developed tool accessible at http://heal.covenantuniversity.edu.ng for conditional essentiality predictions.

Machine learning on large scale perturbation screens for SARS-CoV-2 host factors identifies ß-catenin/CBP inhibitor PRI-724 as a potent antiviral.

Expanding antiviral treatment options against SARS-CoV-2 remains crucial as the virus evolves under selection pressure which already led to the emergence of several drug resistant strains. Broad spectrum host-directed antivirals (HDA) are promising therapeutic options, however the robust identification of relevant host factors by CRISPR/Cas9 or RNA interference screens remains challenging due to low consistency in the resulting hits. To address this issue, we employed machine learning, based on experimental data from several knockout screens and a drug screen. We trained classifiers using genes essential for virus life cycle obtained from the knockout screens. The machines based their predictions on features describing cellular localization, protein domains, annotated gene sets from Gene Ontology, gene and protein sequences, and experimental data from proteomics, phospho-proteomics, protein interaction and transcriptomic profiles of SARS-CoV-2 infected cells. The models reached a remarkable performance suggesting patterns of intrinsic data consistency. The predicted HDF were enriched in sets of genes particularly encoding development, morphogenesis, and neural processes. Focusing on development and morphogenesis-associated gene sets, we found ß-catenin to be central and selected PRI-724, a canonical ß-catenin/CBP disruptor, as a potential HDA. PRI-724 limited infection with SARS-CoV-2 variants, SARS-CoV-1, MERS-CoV and IAV in different cell line models. We detected a concentration-dependent reduction in cytopathic effects, viral RNA replication, and infectious virus production in SARS-CoV-2 and SARS-CoV-1-infected cells. Independent of virus infection, PRI-724 treatment caused cell cycle deregulation which substantiates its potential as a broad spectrum antiviral. Our proposed machine learning concept supports focusing and accelerating the discovery of host dependency factors and identification of potential host-directed antivirals.

Validation of the SACOV-19 score for identifying patients at risk of complicated or more severe COVID-19: a prospective study.

PURPOSE: Identification of patients at risk of complicated or more severe COVID-19 is of pivotal importance, since these patients might require monitoring, antiviral treatment, and hospitalization. In this study, we prospectively evaluated the SACOV-19 score for its ability to predict complicated or more severe COVID-19. METHODS: In this prospective multicenter study, we included 124 adult patients with acute COVID-19 in three German hospitals, who were diagnosed in an early, uncomplicated stage of COVID-19 within 72 h of inclusion. We determined the SACOV-19 score at baseline and performed a follow-up at 30 days. RESULTS: The SACOV-19 score's AUC was 0.816. At a cutoff of > 3, it predicted deterioration to complicated or more severe COVID-19 with a sensitivity of 94% and a specificity of 55%. It performed significantly better in predicting complicated COVID-19 than the random tree-based SACOV-19 predictive model, the CURB-65, 4C mortality, or qCSI scores. CONCLUSION: The SACOV-19 score is a feasible tool to aid decision making in acute COVID-19.

Neurofilament light chain levels predict encephalopathy and outcome in community-acquired pneumonia.

OBJECTIVE: Serum neurofilament light chain (sNfL) is a biomarker for neuroaxonal damage and has been found to be elevated in several neurological diseases with neuronal destruction. New onset of confusion is a hallmark of severity in infections. The objective of this study was to determine whether sNfL levels are increased in patients with community-acquired pneumonia (CAP) and if increased sNfL levels are associated with disease-associated confusion or disease severity. METHODS: In this observational study, sNfL levels were determined with single-molecule array technology in CAP patients of the CAPNETZ cohort with validated CRB (confusion, respiratory rate, and blood pressure)-65 score. We determined associations between log-transformed sNfL concentrations, well-defined clinical characteristics, and unfavorable outcome in multivariable analyses. Receiver operating characteristic (ROC) analysis was performed to assess the prediction accuracy of sNfL levels for confusion in CAP patients. RESULTS: sNfL concentrations were evaluated in 150 CAP patients. Patients with confusion had higher sNfL levels as compared to non-confusion patients of comparable overall disease severity. ROC analysis of sNfL and confusion provided an area under the curve (AUC) of 0.73 (95% CI 0.62-0.82). Log-transformed sNfL levels were not associated with general disease severity. In a logistic regression analysis, log2-sNfL was identified as a strong predictor for an unfavorable outcome. INTERPRETATION: sNfL levels are specifically associated with confusion and not with pneumonia disease severity, thus reflecting a potential objective marker for encephalopathy in these patients. Furthermore, sNfL levels are also associated with unfavorable outcome in these patients and might help clinicians to identify patients at risk.

Linear programming based gene expression model (LPM-GEM) predicts the carbon source for Bacillus subtilis.

BACKGROUND: Elucidating cellular metabolism led to many breakthroughs in biotechnology, synthetic biology, and health sciences. To date, deriving metabolic fluxes by 13C tracer experiments is the most prominent approach for studying metabolic fluxes quantitatively, often with high accuracy and precision. However, the technique has a high demand for experimental resources. Alternatively, flux balance analysis (FBA) has been employed to estimate metabolic fluxes without labeling experiments. It is less informative but can benefit from the low costs and low experimental efforts and gain flux estimates in experimentally difficult conditions. Methods to integrate relevant experimental data have been emerged to improve FBA flux estimations. Data from transcription profiling is often selected since it is easy to generate at the genome scale, typically embedded by a discretization of differential and non-differential expressed genes coding for the respective enzymes. RESULT: We established the novel method Linear Programming based Gene Expression Model (LPM-GEM). LPM-GEM linearly embeds gene expression into FBA constraints. We implemented three strategies to reduce thermodynamically infeasible loops, which is a necessary prerequisite for such an omics-based model building. As a case study, we built a model of B. subtilis grown in eight different carbon sources. We obtained good flux predictions based on the respective transcription profiles when validating with 13C tracer based metabolic flux data of the same conditions. We could well predict the specific carbon sources. When testing the model on another, unseen dataset that was not used during training, good prediction performance was also observed. Furthermore, LPM-GEM outperformed a well-established model building methods. CONCLUSION: Employing LPM-GEM integrates gene expression data efficiently. The method supports gene expression-based FBA models and can be applied as an alternative to estimate metabolic fluxes when tracer experiments are inappropriate.

PITX1 Is a Regulator of TERT Expression in Prostate Cancer with Prognostic Power.

The current risk stratification in prostate cancer (PCa) is frequently insufficient to adequately predict disease development and outcome. One hallmark of cancer is telomere maintenance. For telomere maintenance, PCa cells exclusively employ telomerase, making it essential for this cancer entity. However, TERT, the catalytic protein component of the reverse transcriptase telomerase, itself does not suit as a prognostic marker for prostate cancer as it is rather low expressed. We investigated if, instead of TERT, transcription factors regulating TERT may suit as prognostic markers. To identify transcription factors regulating TERT, we developed and applied a new gene regulatory modeling strategy to a comprehensive transcriptome dataset of 445 primary PCa. Six transcription factors were predicted as TERT regulators, and most prominently, the developmental morphogenic factor PITX1. PITX1 expression positively correlated with telomere staining intensity in PCa tumor samples. Functional assays and chromatin immune-precipitation showed that PITX1 activates TERT expression in PCa cells. Clinically, we observed that PITX1 is an excellent prognostic marker, as concluded from an analysis of more than 15,000 PCa samples. PITX1 expression in tumor samples associated with (i) increased Ki67 expression indicating increased tumor growth, (ii) a worse prognosis, and (iii) correlated with telomere length.

Prediction of COVID-19 deterioration in high-risk patients at diagnosis: an early warning score for advanced COVID-19 developed by machine learning.

PURPOSE: While more advanced COVID-19 necessitates medical interventions and hospitalization, patients with mild COVID-19 do not require this. Identifying patients at risk of progressing to advanced COVID-19 might guide treatment decisions, particularly for better prioritizing patients in need for hospitalization. METHODS: We developed a machine learning-based predictor for deriving a clinical score identifying patients with asymptomatic/mild COVID-19 at risk of progressing to advanced COVID-19. Clinical data from SARS-CoV-2 positive patients from the multicenter Lean European Open Survey on SARS-CoV-2 Infected Patients (LEOSS) were used for discovery (2020-03-16 to 2020-07-14) and validation (data from 2020-07-15 to 2021-02-16). RESULTS: The LEOSS dataset contains 473 baseline patient parameters measured at the first patient contact. After training the predictor model on a training dataset comprising 1233 patients, 20 of the 473 parameters were selected for the predictor model. From the predictor model, we delineated a composite predictive score (SACOV-19, Score for the prediction of an Advanced stage of COVID-19) with eleven variables. In the validation cohort (n = 2264 patients), we observed good prediction performance with an area under the curve (AUC) of 0.73 ± 0.01. Besides temperature, age, body mass index and smoking habit, variables indicating pulmonary involvement (respiration rate, oxygen saturation, dyspnea), inflammation (CRP, LDH, lymphocyte counts), and acute kidney injury at diagnosis were identified. For better interpretability, the predictor was translated into a web interface. CONCLUSION: We present a machine learning-based predictor model and a clinical score for identifying patients at risk of developing advanced COVID-19.

Use of IFNγ/IL10 Ratio for Stratification of Hydrocortisone Therapy in Patients With Septic Shock.

Large clinical trials testing hydrocortisone therapy in septic shock have produced conflicting results. Subgroups may benefit of hydrocortisone treatment depending on their individual immune response. We performed an exploratory analysis of the database from the international randomized controlled clinical trial Corticosteroid Therapy of Septic Shock (CORTICUS) employing machine learning to a panel of 137 variables collected from the Berlin subcohort comprising 83 patients including demographic and clinical measures, organ failure scores, leukocyte counts and levels of circulating cytokines. The identified theranostic marker was validated against data from a cohort of the Hellenic Sepsis Study Group (HSSG) (n = 246), patients enrolled in the clinical trial of Sodium Selenite and Procalcitonin Guided Antimicrobial Therapy in Severe Sepsis (SISPCT, n = 118), and another, smaller clinical trial (Crossover study, n = 20). In addition, in vitro blood culture experiments and in vivo experiments in mouse models were performed to assess biological plausibility. A low serum IFNγ/IL10 ratio predicted increased survival in the hydrocortisone group whereas a high ratio predicted better survival in the placebo group. Using this marker for a decision rule, we applied it to three validation sets and observed the same trend. Experimental studies in vitro revealed that IFNγ/IL10 was negatively associated with the load of (heat inactivated) pathogens in spiked human blood and in septic mouse models. Accordingly, an in silico analysis of published IFNγ and IL10 values in bacteremic and non-bacteremic patients with the Systemic Inflammatory Response Syndrome supported this association between the ratio and pathogen burden. We propose IFNγ/IL10 as a molecular marker supporting the decision to administer hydrocortisone to patients in septic shock. Prospective clinical studies are necessary and standard operating procedures need to be implemented, particularly to define a generic threshold. If confirmed, IFNγ/IL10 may become a suitable theranostic marker for an urging clinical need.

Essential gene prediction in Drosophila melanogaster using machine learning approaches based on sequence and functional features.

Genes are termed to be essential if their loss of function compromises viability or results in profound loss of fitness. On the genome scale, these genes can be determined experimentally employing RNAi or knockout screens, but this is very resource intensive. Computational methods for essential gene prediction can overcome this drawback, particularly when intrinsic (e.g. from the protein sequence) as well as extrinsic features (e.g. from transcription profiles) are considered. In this work, we employed machine learning to predict essential genes in Drosophila melanogaster. A total of 27,340 features were generated based on a large variety of different aspects comprising nucleotide and protein sequences, gene networks, protein-protein interactions, evolutionary conservation and functional annotations. Employing cross-validation, we obtained an excellent prediction performance. The best model achieved in D. melanogaster a ROC-AUC of 0.90, a PR-AUC of 0.30 and a F1 score of 0.34. Our approach considerably outperformed a benchmark method in which only features derived from the protein sequences were used (P < 0.001). Investigating which features contributed to this success, we found all categories of features, most prominently network topological, functional and sequence-based features. To evaluate our approach we performed the same workflow for essential gene prediction in human and achieved an ROC-AUC = 0.97, PR-AUC = 0.73, and F1 = 0.64. In summary, this study shows that using our well-elaborated assembly of features covering a broad range of intrinsic and extrinsic gene and protein features enabled intelligent systems to predict well the essentiality of genes in an organism.

Reprogramming of macrophages employing gene regulatory and metabolic network models.

Upon exposure to different stimuli, resting macrophages undergo classical or alternative polarization into distinct phenotypes that can cause fatal dysfunction in a large range of diseases, such as systemic infection leading to sepsis or the generation of an immunosuppressive tumor microenvironment. Investigating gene regulatory and metabolic networks, we observed two metabolic switches during polarization. Most prominently, anaerobic glycolysis was utilized by M1-polarized macrophages, while the biosynthesis of inosine monophosphate was upregulated in M2-polarized macrophages. Moreover, we observed a switch in the urea cycle. Gene regulatory network models revealed E2F1, MYC, PPARγ and STAT6 to be the major players in the distinct signatures of these polarization events. Employing functional assays targeting these regulators, we observed the repolarization of M2-like cells into M1-like cells, as evidenced by their specific gene expression signatures and cytokine secretion profiles. The predicted regulators are essential to maintaining the M2-like phenotype and function and thus represent potential targets for the therapeutic reprogramming of immunosuppressive M2-like macrophages.

Expression of CCCTC-binding factor (CTCF) is linked to poor prognosis in prostate cancer.

The chromatin-organizing factor CCCTC-binding factor (CTCF) is involved in transcriptional regulation, DNA-loop formation, and telomere maintenance. To evaluate the clinical impact of CTCF in prostate cancer, we analyzed CTCF expression by immunohistochemistry on a tissue microarray containing 17 747 prostate cancers. Normal prostate tissue showed negative to low CTCF expression, while in prostate cancers, CTCF expression was seen in 7726 of our 12 555 (61.5%) tumors and was considered low in 44.6% and high in 17% of cancers. Particularly, high CTCF expression was significantly associated with the presence of the transmembrane protease, serine 2:ETS-related gene fusion: Only 10% of ERG-negative cancers, but 30% of ERG-positive cancers had high-level CTCF expression (P < 0.0001). CTCF expression was significantly associated with advanced pathological tumor stage, high Gleason grade (P < 0.0001 each), nodal metastasis (P = 0.0122), and early biochemical recurrence (P < 0.0001). Multivariable modeling revealed that the prognostic impact of CTCF was independent from established presurgical parameters such as clinical stage and Gleason grade of the biopsy. Comparison with key molecular alterations showed strong associations with the expression of the Ki-67 proliferation marker and presence of phosphatase and tensin homolog deletions (P < 0.0001 each). The results of our study identify CTCF expression as a candidate biomarker for prognosis assessment in prostate cancer.

Macrolide combination therapy for patients hospitalised with community-acquired pneumonia? An individualised approach supported by machine learning.

BACKGROUND: The role of macrolide/ß-lactam combination therapy in community-acquired pneumonia (CAP) of moderate severity is a matter of debate. Macrolides expand the coverage to atypical pathogens and attenuate pulmonary inflammation, but have been associated with cardiovascular toxicity and drug interactions. We developed a decision tree based on aetiological and clinical parameters, which are available ex ante to support a personalised decision for or against macrolides for the best clinical outcome of the individual patient. METHODS: We employed machine learning in a cross-validation scheme based on a well-balanced selection of 4898 patients after propensity score matching to data available on admission of 6440 hospitalised patients with moderate severity (non-intensive care unit patients) from the observational, prospective, multinational CAPNETZ study. We aimed to improve the primary outcome of 180-day survival. RESULTS: We found a simple decision tree of patient characteristics comprising chronic cardiovascular and chronic respiratory comorbidities as well as leukocyte counts in the respiratory secretion at enrolment. Specifically, we found that patients without cardiovascular or patients with respiratory comorbidities and high leukocyte counts in the respiratory secretion benefit from macrolide treatment. Patients identified to be treated in compliance with our treatment suggestion had a lower mortality of 27% (OR 1.83, 95% CI 1.48-2.27; p<0.001) compared to the observed standard of care. CONCLUSION: Stratifying macrolide treatment in patients following a simple treatment rule may lead to considerably reduced mortality in CAP. A future randomised controlled trial confirming our result is necessary before implementing this rule into the clinical routine.

Modelling TERT regulation across 19 different cancer types based on the MIPRIP 2.0 gene regulatory network approach.

BACKGROUND: Reactivation of the telomerase reverse transcriptase gene TERT is a central feature for unlimited proliferation of the majority of cancers. However, the underlying regulatory processes are only partly understood. RESULTS: We assembled regulator binding information from serveral sources to construct a generic human and mouse gene regulatory network. Advancing our "Mixed Integer linear Programming based Regulatory Interaction Predictor" (MIPRIP) approach, we identified the most common and cancer-type specific regulators of TERT across 19 different human cancers. The results were validated by using the well-known TERT regulation by the ETS1 transcription factor in a subset of melanomas with mutations in the TERT promoter. Our improved MIPRIP2 R-package and the associated generic regulatory networks are freely available at https://github.com/KoenigLabNM/MIPRIP. CONCLUSION: MIPRIP 2.0 identified common as well as tumor type specific regulators of TERT. The software can be easily applied to transcriptome datasets to predict gene regulation for any gene and disease/condition under investigation.

Loss of CCAAT-enhancer-binding protein alpha (CEBPA) is linked to poor prognosis in PTEN deleted and TMPRSS2:ERG fusion type prostate cancers.

BACKGROUND: The transcription factor CCAAT-enhancer-binding protein alpha (CEBPA) is a crucial regulator of cell proliferation and differentiation. Expression levels of CEBPA have been suggested to be prognostic in various tumor types. METHODS: Here, we analyzed the immunohistochemical expression of CEBPA in a tissue microarray containing more than 17 000 prostate cancer specimens with annotated clinical and molecular data including for example TMPRSS2:ERG fusion and PTEN deletion status. RESULTS: Normal prostate glands showed moderate to strong CEBPA staining, while CEBPA expression was frequently reduced (40%) or lost (30%) in prostate cancers. Absence of detectable CEBPA expression was markedly more frequent in ERG negative (45%) as compared to ERG positive cancers (20%, P < 0.0001). Reduced CEBPA expression was linked to unfavorable phenotype (P < 0.0001) and poor prognosis (P = 0.0008). Subgroup analyses revealed, that the prognostic value of CEBPA loss was entirely driven by tumors carrying both TMPRSS2:ERG fusions and PTEN deletions. In this subgroup, CEBPA loss was tightly linked to advanced tumor stage (P < 0.0001), high Gleason grade (P < 0.0001), positive nodal stage (0.0003), and early biochemical recurrence (P = 0.0007), while these associations were absent or markedly diminished in tumors with normal PTEN copy numbers and/or absence of ERG fusion. CONCLUSIONS: CEBPA is down regulated in about one third of prostate cancers, but the clinical impact of CEBPA loss is strictly limited to the subset of about 10% prostate cancers carrying both ERG fusion and deletions of the PTEN tumor suppressor. Our findings challenge the concept that prognostic molecular markers may be generally applicable to all prostate cancers.

MiR-192, miR-200c and miR-17 are fibroblast-mediated inhibitors of colorectal cancer invasion.

Colorectal cancer remains a leading cause of cancer-related death worldwide. A previous transcriptomics based study characterized molecular subgroups of which the stromal subgroup was associated with the worst clinical outcome. Micro-RNAs (miRNAs) are well-known regulators of gene expression and can follow a non-linear repression mechanism. We set up a model combining piecewise linear and linear regression and applied this combined regression model to a comprehensive colon adenocarcinoma dataset. We identified miRNAs involved in regulating characteristic gene sets, particularly extracellular matrix remodeling in the stromal subgroup. Comparison of expression data from separated (epithelial) cancer cells and stroma cells or fibroblasts associate these regulatory interactions with infiltrating stromal or tumor-associated fibroblasts. MiR-200c, miR-17 and miR-192 were identified as the most promising candidates regulating genes crucial for extracellular matrix remodeling. We validated our computational findings by in vitro assays. Enforced expression of either miR-200c, miR-17 or miR-192 in untransformed human colon fibroblasts down-regulated 85% of all predicted target genes. Expressing these miRNAs singly or in combination in human colon fibroblasts co-cultured with colon cancer cells considerably reduced cancer cell invasion validating these miRNAs as cancer cell infiltration suppressors in tumor associated fibroblasts.

Fungal biomarker discovery by integration of classifiers.

BACKGROUND: The human immune system is responsible for protecting the host from infection. However, in immunocompromised individuals the risk of infection increases substantially with possible drastic consequences. In extreme, systemic infection can lead to sepsis which is responsible for innumerous deaths worldwide. Amongst its causes are infections by bacteria and fungi. To increase survival, it is mandatory to identify the type of infection rapidly. Discriminating between fungal and bacterial pathogens is key to determine if antifungals or antibiotics should be administered, respectively. For this, in situ experiments have been performed to determine regulation mechanisms of the human immune system to identify biomarkers. However, these studies led to heterogeneous results either due different laboratory settings, pathogen strains, cell types and tissues, as well as the time of sample extraction, to name a few. METHODS: To generate a gene signature capable of discriminating between fungal and bacterial infected samples, we employed Mixed Integer Linear Programming (MILP) based classifiers on several datasets comprised of the above mentioned pathogens. RESULTS: When combining the classifiers by a joint optimization we could increase the consistency of the biomarker gene list independently of the experimental setup. An increase in pairwise overlap (the number of genes that overlap in each cross-validation) of 43% was obtained by this approach when compared to that of single classifiers. The refined gene list was composed of 19 genes and ranked according to consistency in expression (up- or down-regulated) and most of them were linked either directly or indirectly to the ERK-MAPK signalling pathway, which has been shown to play a key role in the immune response to infection. Testing of the identified 12 genes on an unseen dataset yielded an average accuracy of 83%. CONCLUSIONS: In conclusion, our method allowed the combination of independent classifiers and increased consistency and reliability of the generated gene signatures.

SOX5 is involved in balanced MITF regulation in human melanoma cells.

BACKGROUND: Melanoma is a cancer with rising incidence and new therapeutics are needed. For this, it is necessary to understand the molecular mechanisms of melanoma development and progression. Melanoma differs from other cancers by its ability to produce the pigment melanin via melanogenesis; this biosynthesis is essentially regulated by microphthalmia-associated transcription factor (MITF). MITF regulates various processes such as cell cycling and differentiation. MITF shows an ambivalent role, since high levels inhibit cell proliferation and low levels promote invasion. Hence, well-balanced MITF homeostasis is important for the progression and spread of melanoma. Therefore, it is difficult to use MITF itself for targeted therapy, but elucidating its complex regulation may lead to a promising melanoma-cell specific therapy. METHOD: We systematically analyzed the regulation of MITF with a novel established transcription factor based gene regulatory network model. Starting from comparative transcriptomics analysis using data from cells originating from nine different tumors and a melanoma cell dataset, we predicted the transcriptional regulators of MITF employing ChIP binding information from a comprehensive set of databases. The most striking regulators were experimentally validated by functional assays and an MITF-promoter reporter assay. Finally, we analyzed the impact of the expression of the identified regulators on clinically relevant parameters of melanoma, i.e. the thickness of primary tumors and patient overall survival. RESULTS: Our model predictions identified SOX10 and SOX5 as regulators of MITF. We experimentally confirmed the role of the already well-known regulator SOX10. Additionally, we found that SOX5 knockdown led to MITF up-regulation in melanoma cells, while double knockdown with SOX10 showed a rescue effect; both effects were validated by reporter assays. Regarding clinical samples, SOX5 expression was distinctively up-regulated in metastatic compared to primary melanoma. In contrast, survival analysis of melanoma patients with predominantly metastatic disease revealed that low SOX5 levels were associated with a poor prognosis. CONCLUSION: MITF regulation by SOX5 has been shown only in murine cells, but not yet in human melanoma cells. SOX5 has a strong inhibitory effect on MITF expression and seems to have a decisive clinical impact on melanoma during tumor progression.

Mixed Integer Linear Programming based machine learning approach identifies regulators of telomerase in yeast.

Understanding telomere length maintenance mechanisms is central in cancer biology as their dysregulation is one of the hallmarks for immortalization of cancer cells. Important for this well-balanced control is the transcriptional regulation of the telomerase genes. We integrated Mixed Integer Linear Programming models into a comparative machine learning based approach to identify regulatory interactions that best explain the discrepancy of telomerase transcript levels in yeast mutants with deleted regulators showing aberrant telomere length, when compared to mutants with normal telomere length. We uncover novel regulators of telomerase expression, several of which affect histone levels or modifications. In particular, our results point to the transcription factors Sum1, Hst1 and Srb2 as being important for the regulation of EST1 transcription, and we validated the effect of Sum1 experimentally. We compiled our machine learning method leading to a user friendly package for R which can straightforwardly be applied to similar problems integrating gene regulator binding information and expression profiles of samples of e.g. different phenotypes, diseases or treatments.

Characterizing protein interactions employing a genome-wide siRNA cellular phenotyping screen.

Characterizing the activating and inhibiting effect of protein-protein interactions (PPI) is fundamental to gain insight into the complex signaling system of a human cell. A plethora of methods has been suggested to infer PPI from data on a large scale, but none of them is able to characterize the effect of this interaction. Here, we present a novel computational development that employs mitotic phenotypes of a genome-wide RNAi knockdown screen and enables identifying the activating and inhibiting effects of PPIs. Exemplarily, we applied our technique to a knockdown screen of HeLa cells cultivated at standard conditions. Using a machine learning approach, we obtained high accuracy (82% AUC of the receiver operating characteristics) by cross-validation using 6,870 known activating and inhibiting PPIs as gold standard. We predicted de novo unknown activating and inhibiting effects for 1,954 PPIs in HeLa cells covering the ten major signaling pathways of the Kyoto Encyclopedia of Genes and Genomes, and made these predictions publicly available in a database. We finally demonstrate that the predicted effects can be used to cluster knockdown genes of similar biological processes in coherent subgroups. The characterization of the activating or inhibiting effect of individual PPIs opens up new perspectives for the interpretation of large datasets of PPIs and thus considerably increases the value of PPIs as an integrated resource for studying the detailed function of signaling pathways of the cellular system of interest.

Estimating the activity of transcription factors by the effect on their target genes.

MOTIVATION: Understanding regulation of transcription is central for elucidating cellular regulation. Several statistical and mechanistic models have come up the last couple of years explaining gene transcription levels using information of potential transcriptional regulators as transcription factors (TFs) and information from epigenetic modifications. The activity of TFs is often inferred by their transcription levels, promoter binding and epigenetic effects. However, in principle, these methods do not take hard-to-measure influences such as post-transcriptional modifications into account. RESULTS: For TFs, we present a novel concept circumventing this problem. We estimate the regulatory activity of TFs using their cumulative effects on their target genes. We established our model using expression data of 59 cell lines from the National Cancer Institute. The trained model was applied to an independent expression dataset of melanoma cells yielding excellent expression predictions and elucidated regulation of melanogenesis. AVAILABILITY AND IMPLEMENTATION: Using mixed-integer linear programming, we implemented a switch-like optimization enabling a constrained but optimal selection of TFs and optimal model selection estimating their effects. The method is generic and can also be applied to further regulators of transcription. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.