RESUMO
Sparse coding can improve discrimination of sensory stimuli by reducing overlap between their representations. Two factors, however, can offset sparse coding's benefits: similar sensory stimuli have significant overlap and responses vary across trials. To elucidate the effects of these 2 factors, we analyzed odor responses in the fly and mouse olfactory regions implicated in learning and discrimination-the mushroom body (MB) and the piriform cortex (PCx). We found that neuronal responses fall along a continuum from extremely reliable across trials to extremely variable or stochastic. Computationally, we show that the observed variability arises from noise within central circuits rather than sensory noise. We propose this coding scheme to be advantageous for coarse- and fine-odor discrimination. More reliable cells enable quick discrimination between dissimilar odors. For similar odors, however, these cells overlap and do not provide distinguishing information. By contrast, more unreliable cells are decorrelated for similar odors, providing distinguishing information, though these benefits only accrue with extended training with more trials. Overall, we have uncovered a conserved, stochastic coding scheme in vertebrates and invertebrates, and we identify a candidate mechanism, based on variability in a winner-take-all (WTA) inhibitory circuit, that improves discrimination with training.
Assuntos
Dípteros , Percepção Olfatória , Animais , Camundongos , Condutos Olfatórios/fisiologia , Olfato/fisiologia , Odorantes , Aprendizagem/fisiologia , Percepção Olfatória/fisiologiaRESUMO
Tyrosine kinase inhibitors (TKIs) have drastically improved the outcomes of pCML (paediatric CML) but data on long-term off-target toxicities of TKIs in children are scarce. In this single-centre, retrospective cum prospective study of pCML in chronic phase, we report our experience of treating 173 children with imatinib and following them for long-term toxicities. Mean (SD) time to attain CHR, CCyR and MMR were 3.05 (2.1), 10.6 (8.4) and 43.4 (31.8) months respectively. DMR was not attained in 59 (34%) patients at last follow-up. Ten patients were switched to second-generation TKIs (2G-TKIs; nilotinib = 1/dasatinib = 9) due to poor/loss in response, of which seven had kinase domain mutations. Three patients progressed to the blastic phase. At a median follow-up of 84 (3-261) months, the 5-year EFS and OS for the entire cohort were 96.9% (95% CI: 93.4-100) and 98.7% (95% CI: 96.9-100) respectively. Screening for long-term toxicities revealed low bone density and hypovitaminosis D in 70% and 80% respectively. Other late effects included short stature (27%), delayed puberty (15%), poor sperm quality (43%) and miscellaneous endocrinopathies (8%). Children younger than 5 years at diagnosis were more susceptible to growth and endocrine toxicities (p = 0.009). Regular monitoring for long-term toxicities, timely intervention and trial of discontinuation whenever feasible are likely to improve the long-term outlook of pCML.
Assuntos
Leucemia Mielogênica Crônica BCR-ABL Positiva , Criança , Humanos , Masculino , Dasatinibe , Seguimentos , Hospitais , Leucemia Mielogênica Crônica BCR-ABL Positiva/genética , Estudos Prospectivos , Inibidores de Proteínas Quinases/efeitos adversos , Estudos Retrospectivos , Sêmen , Resultado do Tratamento , Pré-EscolarRESUMO
BACKGROUND: The surveillance guidelines following treatment completion for patients with high-grade sarcomas of the extremities are based largely upon expert opinions and consensus. In the current meta-analysis, we aim to study the utility of surveillance imaging to diagnose local and metastatic pulmonary relapses among patients with extremity soft tissue sarcomas and primary bone sarcomas. PATIENTS AND METHODS: A meta-analysis was performed to assess the sensitivity, specificity and diagnostic odds ratio (DOR) of surveillance imaging to diagnose local and metastatic pulmonary relapse among patients with sarcoma of the extremities. In addition, impact of surveillance imaging on overall survival was assessed. Heterogeneity among eligible studies was evaluated by I2 statistics. Sensitivity analysis was assessed using influence plots and Baujat plots. RESULTS: Ten studies including 2160 patients with sarcoma were found eligible. For diagnoses of local recurrence based on surveillance imaging (nine studies, 1917 patients), the estimated sensitivity, specificity, and DOR were 13.6%, 99.5%, and 78.15, respectively. Only 16.7% of local recurrences were diagnosed based on imaging. For diagnoses of metastatic pulmonary recurrence (eight studies; 1868 patients), estimated sensitivity, specificity, and DOR were 76.1%, 99.3%, and 1059.9, respectively. A sensitivity analysis showed significant heterogeneity among included studies. None of the included studies showed an overall-survival benefit with the use of surveillance imaging. CONCLUSION: The current meta-analysis challenges the notion of routine use of imaging to detect local relapse, while favoring chest imaging, using either chest radiography or computed tomography scan, for surveillance. Further studies are required to study the ideal surveillance strategy including timing and imaging modality.
Assuntos
Neoplasias Ósseas , Sarcoma , Neoplasias de Tecidos Moles , Humanos , Recidiva Local de Neoplasia/epidemiologia , Sarcoma/patologia , Neoplasias Ósseas/diagnóstico por imagem , Recidiva , Pulmão/patologia , Extremidades/diagnóstico por imagem , Extremidades/patologia , Neoplasias de Tecidos Moles/patologiaRESUMO
INTRODUCTION: Childhood cancers are a significant global health concern, particularly in low- and middle-income countries (LMICs), where over 80% of childhood cancer patients reside. In India, acute myeloid leukemia (AML) constitutes a significant portion of childhood cancers; however, the data on the cost-effectiveness of childhood AML treatment in India and other LMICs remain limited. METHODS: The study focused on children (<15 years of age) diagnosed with AML at a tertiary care cancer center in North India. Data, including treatment outcome, treatment-related morbidity, mortality, and costs were retrospectively collected from the electronic medical record and hospital database. Cost-effectiveness was assessed using disability-adjusted life years (DALY) averted in relation to the country-specific cost-effectiveness threshold. RESULTS: Among 59 AML patients, treatment-related high mortality rates, abandonment, and limited access to bone marrow transplantation were notable challenges. Intensive chemotherapy resulted in substantial sepsis-related complications, with treatment-related mortality reaching 30%. The 3-year event-free survival and overall survival of the 43 patients who received intensive therapy were 24.5% ± 7.6% and 27.9% ± 8.3%, respectively. Despite these challenges, treating childhood AML was still found to be cost-effective. The total cost per newly diagnosed patient treated with curative intent was $4454. Cost per DALY averted accounted for 24% of the gross domestic product (GDP) per capita, rendering the treatment to be cost-effective with a stringent cost-effectiveness threshold utilized. CONCLUSION: The study underscores the challenges faced while treating childhood AML in LMICs, including treatment-induced high sepsis-related mortality and abandonment. Despite these challenges, it remains cost-effective to treat childhood AML in India. Future efforts should focus on reducing treatment-related morbidity and mortality to further improve outcomes and cost-effectiveness.
RESUMO
BACKGROUND: Pediatric core binding factor acute myeloid leukemia (CBF-AML), although considered a favorable risk subtype, exhibits variable outcomes primarily driven by additional genetic abnormalities, such as KIT mutations. PROCEDURE: In this study, we examined the prognostic impact of KIT mutations in 130 pediatric patients with CBF-AML, treated uniformly at a single center over 4 years (2017-2021). KIT mutations were detected via next-generation sequencing using a myeloid panel comprising 52 genes for most patients. RESULTS: Our findings revealed that KIT mutations were present in 31% of CBF-AML cases. Exon 17 KIT mutation was most commonly (72%) seen with notable occurrences at the D816 and N822 residue in 48% and 39% of cases, respectively. The 3-year cumulative incidence of relapse (CIR) and overall survival (OS) for patients with exon 17 KIT mutation were 36% and 40%, respectively, and was significantly worse in comparison to other site KIT mutations (3-year CIR: 11%; OS: 64%) and without KIT mutation (3-year CIR: 13%; OS:71%). Notably, the prognostic impact of KIT mutations was prominent in patients with RUNX1::RUNX1T1, but not in those with CBFB::MYH11 fusion. Additionally, a high KIT variant-allele frequency (VAF) (>33%) predicted for a higher disease relapse; 3-year CIR of 40% for VAF greater than 33% versus 7% for VAF less than 33%. When adjusted for site of KIT mutation and end-of-induction measurable residual disease, VAF greater than 33% correlated with poor OS (hazard ratio [HR]: 4.4 [95% CI: 1.2-17.2], p = .034). CONCLUSION: Exon 17 KIT mutations serve as an important predictor of relapse in RUNX1::RUNX1T1 pediatric AML. In addition, a high KIT VAF may predict poor outcomes in these patients. These results emphasize the need to incorporate KIT mutational analysis into risk stratification for pediatric CBF-AML.
Assuntos
Subunidade alfa 2 de Fator de Ligação ao Core , Leucemia Mieloide Aguda , Humanos , Criança , Subunidade alfa 2 de Fator de Ligação ao Core/genética , Proteínas Proto-Oncogênicas c-kit/genética , Leucemia Mieloide Aguda/terapia , Mutação , Prognóstico , Éxons/genética , Recidiva , Proteína 1 Parceira de Translocação de RUNX1/genéticaRESUMO
Why the cerebral cortex folds in some mammals but not in others has long fascinated and mystified neurobiologists. Over the past century-especially the past decade-researchers have used theory and experiment to support different folding mechanisms such as tissue buckling from mechanical stress, axon tethering, localized proliferation, and external constraints. In this review, we synthesize these mechanisms into a unifying framework and introduce a hitherto unappreciated mechanism, the radial intercalation of new neurons at the top of the cortical plate, as a likely proximate force for tangential expansion that then leads to cortical folding. The interplay between radial intercalation and various biasing factors, such as local variations in proliferation rate and connectivity, can explain the formation of both random and stereotypically positioned folds.
Assuntos
Córtex Cerebral/anatomia & histologia , Córtex Cerebral/citologia , Neurogênese , Animais , Modelos NeurológicosRESUMO
IKZF1 (IKAROS family Zinc Finger 1) alteration is an essential regulator of both T- and B-cell lineage specification with leukemogenic potential. IKZF1 deletion have been described in childhood acute lymphoblastic leukemia (ALL) with varying prevalence often influenced by underlying cytogenetics and also shown to have diverse prognostic significance. We aimed to evaluate the prevalence and prognostic significance of IKZF1 deletion among childhood ALL. Electronic databases of MEDLINE, EMBASE and SCOPUS were searched and 32 studies found eligible. Estimated prevalence of IKZF1 deletion among BCR::ABL1 negative and BCR::ABL1 positive ALL patients was 14% (95%CI:13-16%, I2 = 79%; 26 studies) and 63% (95%CI:59-68% I2 = 42%; 10 studies) respectively. Most common site of IKZF1 deletion was whole chromosome (exon1-8) deletion in 32.3% (95%CI: 23.8-40.7%) followed by exon 4-7 deletion in 28.6% (95%CI: 19.7-37.5%). A positive minimal residual disease at the end of induction was more common among patients with IKZF1 deletion, odds ratio: 3.09 (95%CI:2.3-4.16, I2 = 54%; 15 studies). Event-free survival and overall survival were significantly worse for IKZF1 deletion, hazard ratio (HR): 2.10 (95%CI:1.90-2.32, I2 = 28%; 31 studies) and HR: 2.38 (95%CI:1.93-2.93, I2 = 40; 15 studies) respectively. In summary, the current meta-analysis highlights the frequency of IKZF1 deletion and its negative impact on survival in childhood ALL. Further studies exploring the influence of IKZF1 deletion in the presence of classical cytogenetic and other copy number alterations would further help in characterising its prognostic role.
Assuntos
Leucemia-Linfoma Linfoblástico de Células Precursoras B , Leucemia-Linfoma Linfoblástico de Células Precursoras , Criança , Humanos , Prognóstico , Prevalência , Fator de Transcrição Ikaros/genética , Leucemia-Linfoma Linfoblástico de Células Precursoras/epidemiologia , Leucemia-Linfoma Linfoblástico de Células Precursoras/genética , Intervalo Livre de Progressão , Deleção de Genes , Leucemia-Linfoma Linfoblástico de Células Precursoras B/genéticaRESUMO
The outlook of relapsed ALL in low- and middle-income countries (LMICs) is dismal due to high treatment-related toxicities and inadequate resources. We report our experience of using a locally adapted mitoxantrone-based protocol for non-high risk (HR) relapsed B-ALL (rALL). A retrospective cum prospective study of standard and intermediate risk (SR and IR) rALL patients treated on TMH rALL-18 protocol (adapted from COG/UKALLR3/Int-Re-ALL protocols) between November 2018 and January 2021 was analyzed. The protocol comprising of 7 blocks of multi-agent chemotherapy including mitoxantrone in induction followed by local irradiation and maintenance, underwent serial modifications based on our experience with initial patients. Eighty-two patients (SR rALL, 3; IR rALL, 79) were treated on TMH rALL-18 protocol. Of 321 grade 3/4 reported toxicities, around 43% (138 toxicities) were noted during induction. Induction chemotherapy was outpatient-based; however, 68 patients (82.9%) required supportive care admissions. Twelve out of 19 patients with gram negative bacilli sepsis (included 7 MDRO) died during reinduction. Five remission deaths were seen during block 3 after which cytarabine was dose reduced (3 g to 2 g/m2). Post-reinduction minimal residual disease was negative in 54 (80.6%) out of 67 evaluable patients. At a median follow-up of 24 months (95% CI 22-27), the estimated 2-year event-free and overall survival of the entire cohort was 58% (95% CI 48.1-69.9) and 60.3% (95% CI 50.5-72). Until the time, targeted therapies are freely accessible in LMICs, strengthening supportive care as well as local adaptation of protocols that strike a fine balance between efficacy and tolerability are mandated.
Assuntos
Bacteriemia , Mitoxantrona , Humanos , Criança , Estudos Prospectivos , Estudos Retrospectivos , Hospitais , Índia/epidemiologiaRESUMO
Sepsis-related mortality continues to be a major concern while treating pediatric cancer patients, more so with the rise in the incidence of multidrug-resistant organisms (MDRO). In this retrospective study conducted between January 2021 and December 2022 at a tertiary cancer center in India, granulocyte transfusion was offered in addition to standard antimicrobial therapy to 64 children with hematolymphoid malignancy who developed 75 episodes of severe sepsis following intensive chemotherapy. Forty-four (83%) of 53 blood culture proven sepsis was caused by MDROs. Thirty-seven (70%) patients with blood culture proven sepsis cleared the organism after granulocyte transfusion. Thirty-day mortality was 25% for the entire study cohort and 32% for patients with MDRO sepsis.
RESUMO
BACKGROUND: While anthracycline therapy has been shown to improve outcomes in Ewing sarcoma, it may be associated with severe and even fatal cardiac dysfunction. We evaluated the burden and determinants of cardiac dysfunction in pediatric Ewing sarcoma (pES). METHODS: This retrospective study included children aged 0-18 years with pES treated at our center with the EFT 2001 protocol (anthracycline and cyclophosphamide containing regimen), with/without radiation therapy from January 2001 to December 2018. Cardiac dysfunction was defined as left ventricular (LV) ejection fraction with an absolute value <50%. RESULTS: Amongst 650 eligible patients (median age at diagnosis 12 years and median follow-up duration 69 months), 85 (13%) developed cardiac dysfunction, at a median 13 months (range: 1-168 months). The cumulative incidence of cardiac dysfunction was 5.7% at 12 months, 12% at 2 years, 13% at 3 years, 14% at 5 years, and 15 % at 10 years. At a median follow-up duration of 25 (range: 3-212) months, 21 (24.7%) patients had normalization of LV function, whereas nine (10.6%) patients died of cardiac causes. Older age at diagnosis (7-12 years OR 5.1, p = .01, 13-18 years, OR 3.9, p = .03), female sex (OR 2.3, p = .004), undernutrition (OR 2.9, p = .001), and chest wall location (OR 8.7, p = .08) were risk factors for cardiac dysfunction. CONCLUSIONS: Children with Ewing sarcoma have a high incidence of cardiac dysfunction, which continues to develop even years after therapy, underlining the need for life-long surveillance. Undernourished children are at a higher risk for cardiac dysfunction and need stringent monitoring.
RESUMO
BACKGROUND: Not all the significant progress made in the management of children with hepatoblastoma (HB) has translated into improved outcomes in limited-resource settings. There are limited data on outcomes in children with HB from India. METHODS: All patients diagnosed with HB between July 2013 and December 2020 were risk-stratified and treated as per International Liver Tumor Strategy Group (SIOPEL). Patients with standard-risk HB received cisplatin monotherapy and those with high-risk HB received alternating cycles of cisplatin and the combination of carboplatin plus doxorubicin. Data regarding demographic details, chemotherapy, surgery, liver transplantation, outcomes, prognostic factors, and toxicity were collected. RESULTS: Of 157 patients with HB, 117 (74%) were high risk, 31 (20%) were standard risk, and nine (6%) unknown. Patients with standard-risk disease had excellent outcomes, with 3-year event-free survival (EFS) and overall survival (OS) of 96% and 100%, respectively. Among high-risk HB, six underwent orthotopic liver transplantation of which four were alive at last follow-up. The 3-year EFS and OS of patients with high-risk disease was 56% and 66%, respectively. Outcomes of patients with PRETEXT IV (3-year EFS: 42%, 3-year OS: 50%) and metastatic disease (3-year EFS: 30%, 3-year OS: 50%) were dismal. Patients with serum alpha-fetoprotein (AFP) reduction greater than 90% following two courses of chemotherapy had favorable outcomes; 3-year EFS: 80% versus 58% (p = .013) and 3-year OS: 95% vs. 68% (p < .01). Only two (6%) of 31 patients with relapse/refractory HB were alive at a median follow-up of 36 months, and both had received salvage chemotherapy and surgery. CONCLUSIONS: While children with standard-risk HB had excellent outcomes, those with high-risk disease continue to do poorly. Serial monitoring of serum AFP values is a cost-effective and reliable predictor of outcomes. Orthotopic liver transplantation remains a viable option for inoperable disease in resource-limited settings as well.
Assuntos
Hepatoblastoma , Neoplasias Hepáticas , Criança , Humanos , Lactente , Hepatoblastoma/patologia , Cisplatino , Prognóstico , alfa-Fetoproteínas , Protocolos de Quimioterapia Combinada Antineoplásica/uso terapêutico , Recidiva Local de Neoplasia/patologia , Neoplasias Hepáticas/patologia , Resultado do Tratamento , Carboplatina , DoxorrubicinaRESUMO
MOTIVATION: In kinetic models of metabolism, the parameter values determine the dynamic behaviour predicted by these models. Estimating parameters from in vivo experimental data require the parameters to be structurally identifiable, and the data to be informative enough to estimate these parameters. Existing methods to determine the structural identifiability of parameters in kinetic models of metabolism can only be applied to models of small metabolic networks due to their computational complexity. Additionally, a priori experimental design, a necessity to obtain informative data for parameter estimation, also does not account for using steady-state data to estimate parameters in kinetic models. RESULTS: Here, we present a scalable methodology to structurally identify parameters for each flux in a kinetic model of metabolism based on the availability of steady-state data. In doing so, we also address the issue of determining the number and nature of experiments for generating steady-state data to estimate these parameters. By using a small metabolic network as an example, we show that most parameters in fluxes expressed by mechanistic enzyme kinetic rate laws can be identified using steady-state data, and the steady-state data required for their estimation can be obtained from selective experiments involving both substrate and enzyme level perturbations. The methodology can be used in combination with other identifiability and experimental design algorithms that use dynamic data to determine the most informative experiments requiring the least resources to perform. AVAILABILITY AND IMPLEMENTATION: https://github.com/LMSE/ident. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Assuntos
Modelos Biológicos , Projetos de Pesquisa , Algoritmos , Cinética , Redes e Vias MetabólicasRESUMO
BACKGROUND: Even though rituximab has emerged as standard of care for the management of high-risk pediatric Burkitt lymphoma (BL), its safety in children from the low-middle-income countries (LMICs) remains to be proven. We herein report our experience of using rituximab in children with BL. METHODS: All patients diagnosed with BL between January 2015 and December 2017 were treated in a risk-stratified manner with either the modified MCP-842 or modified LMB protocol. Patients with poor response to MCP-842 were switched to the LMB-salvage regimen. In addition, rituximab was given to selected high-risk patients. RESULT: Forty-two (49.4%) of 85 patients with BL received rituximab. The incidence of febrile neutropenia (90.5% vs 67.4%; P = 0.02), pneumonia (38.1% vs 11.6%; P = 0.005), intensive care unit admissions (54.5% vs 17.6%; P = 0.002), and toxic deaths (26.2% vs 9.3%; P = 0.04) was higher among BL patients who received rituximab. Pneumonia was fatal in 11 of 16 (69%) patients who received rituximab. On multivariate analysis, rituximab continued to be significantly associated with toxic deaths ( OR: 11.45 [95% CI: 1.87-70.07; P = 0.008]). The addition of rituximab to intensive chemotherapy resulted in an inferior one-year event-free survival (49.4% ± 8.1% vs 79.3% ± 6.5%; P = 0.025) and one-year overall survival (63.1% ± 8.5% vs 91.8% ± 4.5%; P = 0.007) with no improvement in one-year relapse-free survival (78.3% ± 7.3% vs 83.9% ± 6.0%; P = 0.817). CONCLUSION: Rituximab was associated with increased toxicities and toxic deaths in our patients. The potential immunomodulatory effect of rituximab and increased susceptibility to infections in patients from LMICs have to be carefully considered while choosing this drug in the treatment of BL in resource-constrained settings.
Assuntos
Antineoplásicos Imunológicos/efeitos adversos , Linfoma de Burkitt/tratamento farmacológico , Efeitos Colaterais e Reações Adversas Relacionados a Medicamentos/etiologia , Rituximab/efeitos adversos , Adolescente , Criança , Pré-Escolar , Efeitos Colaterais e Reações Adversas Relacionados a Medicamentos/patologia , Feminino , Seguimentos , Humanos , Masculino , Prognóstico , Estudos Retrospectivos , Taxa de SobrevidaRESUMO
Animals successfully thrive in noisy environments with finite resources. The necessity to function with resource constraints has led evolution to design animal brains (and bodies) to be optimal in their use of computational power while being adaptable to their environmental niche. A key process undergirding this ability to adapt is the process of learning. Although a complete characterization of the neural basis of learning remains ongoing, scientists for nearly a century have used the brain as inspiration to design artificial neural networks capable of learning, a case in point being deep learning. In this viewpoint, we advocate that deep learning can be further enhanced by incorporating and tightly integrating five fundamental principles of neural circuit design and function: optimizing the system to environmental need and making it robust to environmental noise, customizing learning to context, modularizing the system, learning without supervision, and learning using reinforcement strategies. We illustrate how animals integrate these learning principles using the fruit fly olfactory learning circuit, one of nature's best-characterized and highly optimized schemes for learning. Incorporating these principles may not just improve deep learning but also expose common computational constraints. With judicious use, deep learning can become yet another effective tool to understand how and why brains are designed the way they are.
Assuntos
Aprendizado Profundo , Modelos Neurológicos , Vias Aferentes/fisiologia , Animais , Aprendizagem por Associação/fisiologia , Aprendizagem da Esquiva/fisiologia , Condicionamento Clássico/fisiologia , Condicionamento Operante/fisiologia , Drosophila melanogaster/fisiologia , Meio Ambiente , Corpos Pedunculados/fisiologia , Rede Nervosa/fisiologia , Neurópilo/fisiologia , Odorantes , Percepção Olfatória/fisiologia , Neurônios Receptores Olfatórios/fisiologia , Reforço Psicológico , Recompensa , Detecção de Sinal Psicológico/fisiologia , Razão Sinal-Ruído , Sinapses/fisiologiaRESUMO
Motivation: Metabolism can exhibit dynamic phenomena like bistability due to the presence of regulatory motifs like the positive feedback loop. As cell factories, microorganisms with bistable metabolism can have a high and a low product flux at the two stable steady states, respectively. The exclusion of metabolic regulation and network dynamics limits the ability of pseudo-steady state stoichiometric models to detect the presence of bistability, and reliably assess the outcomes of design perturbations to metabolic networks. Results: Using kinetic models of metabolism, we assess the change in the bistable characteristics of the network, and suggest designs based on perturbations to the positive feedback loop to enable the network to produce at its theoretical maximum rate. We show that the most optimal production design in parameter space, for a small bistable metabolic network, may exist at the boundary of the bistable region separating it from the monostable region of low product fluxes. The results of our analysis can be broadly applied to other bistable metabolic networks with similar positive feedback network topologies. This can complement existing model-based design strategies by providing a smaller number of feasible designs that need to be tested in vivo. Availability and implementation: http://lmse.biozone.utoronto.ca/downloads/. Contact: krishna.mahadevan@utoronto.ca. Supplementary information: Supplementary data are available at Bioinformatics online.