Diagnosing relapse in children's brain tumors using metabolite profiles

BACKGROUND

Malignant brain tumors in children generally have a very poor prognosis when they relapse and improvements are required in their management. It can be difficult to accurately diagnose abnormalities detected during tumor surveillance, and new techniques are required to aid this process. This study investigates how metabolite profiles measured noninvasively by (1)H magnetic resonance spectroscopy (MRS) at relapse reflect those at diagnosis and may be used in this monitoring process.

METHODS

Single-voxel MRS (1.5 T, point-resolved spectroscopy, echo time 30 ms, repetition time 1500 ms was performed on 19 children with grades II-IV brain tumors during routine MRI scans prior to treatment for a suspected brain tumor and at suspected first relapse. MRS was analyzed using TARQUIN software to provide metabolite concentrations. Paired Student's t-tests were performed between metabolite profiles at diagnosis and at first relapse.

RESULTS

There was no significant difference (P > .05) in the level of any metabolite, lipid, or macromolecule from tumors prior to treatment and at first relapse. This was true for the whole group (n = 19), those with a local relapse (n = 12), and those with a distant relapse (n = 7). Lipids at 1.3 ppm were close to significance when comparing the level at diagnosis with that at distant first relapse (P = .07, 6.5 vs 12.9). In 5 cases the MRS indicative of tumor preceded a formal diagnosis of relapse.

CONCLUSIONS

Tumor metabolite profiles, measured by MRS, do not change greatly from diagnosis to first relapse, and this can aid the confirmation of the presence of tumor.

Using evolutional properties of gene networks in understanding survival prognosis of glioblastoma

Previously, we investigated survival prognosis of glioblastoma by applying a gene regulatory approach to a human glioblastoma dataset. Here, we further extend our understanding of survival prognosis of glioblastoma by refining the network inference technique we apply to the glioblastoma dataset with the intent of uncovering further topological properties of the networks. For this study, we modify the approach by specifically looking at both positive and negative correlations separately, as opposed to absolute correlations. There is great interest in applying mathematical modeling approaches to cancer cell line datasets to generate network models of gene regulatory interactions. Analysis of these networks using graph theory metrics can identify genes of interest. The principal approach for modeling microarray datasets has been to group all the cell lines together into one overall network, and then, analyze this network as a whole. As per the previous study, we categorize a human glioblastoma cell line dataset into five categories based on survival data, and analyze each category separately using both negative and positive correlation networks constructed using a modified version of the WGCNA algorithm. Using this approach, we identified a number of genes as being important across different survival stages of the glioblastoma cell lines.

A unified structural/terminological interoperability framework based on LexEVS: application to TRANSFoRm

OBJECTIVE

Biomedical research increasingly relies on the integration of information from multiple heterogeneous data sources. Despite the fact that structural and terminological aspects of interoperability are interdependent and rely on a common set of requirements, current efforts typically address them in isolation. We propose a unified ontology-based knowledge framework to facilitate interoperability between heterogeneous sources, and investigate if using the LexEVS terminology server is a viable implementation method.

MATERIALS AND METHODS

We developed a framework based on an ontology, the general information model (GIM), to unify structural models and terminologies, together with relevant mapping sets. This allowed a uniform access to these resources within LexEVS to facilitate interoperability by various components and data sources from implementing architectures.

RESULTS

Our unified framework has been tested in the context of the EU Framework Program 7 TRANSFoRm project, where it was used to achieve data integration in a retrospective diabetes cohort study. The GIM was successfully instantiated in TRANSFoRm as the clinical data integration model, and necessary mappings were created to support effective information retrieval for software tools in the project.

CONCLUSIONS

We present a novel, unifying approach to address interoperability challenges in heterogeneous data sources, by representing structural and semantic models in one framework. Systems using this architecture can rely solely on the GIM that abstracts over both the structure and coding. Information models, terminologies and mappings are all stored in LexEVS and can be accessed in a uniform manner (implementing the HL7 CTS2 service functional model). The system is flexible and should reduce the effort needed from data sources personnel for implementing and managing the integration.

Increased unsaturation of lipids in cytoplasmic lipid droplets in DAOY cancer cells in response to cisplatin treatment

Increases in ¹H nuclear magnetic resonance spectroscopy (NMR) visible lipids are a well-documented sign of treatment response in cancers. Lipids in cytoplasmic lipid droplets (LDs) are the main contributors to the NMR lipid signals. Two human primitive neuroectodermal tumour cell lines with different sensitivities to cisplatin treatment were studied. Increases in NMR visible saturated and unsaturated lipids in cisplatin treated DAOY cells were associated with the accumulation of LDs prior to DNA fragmentation due to apoptosis. An increase in unsaturated fatty acids (UFAs) was detected in isolated LDs from DAOY cells, in contrast to a slight decrease in UFAs in lipid extracts from whole cells. Oleic acid and linoleic acid were identified as the accumulating UFAs in LDs by heteronuclear single quantum coherence spectroscopy (HSQC). ¹H NMR lipids in non-responding PFSK-1 cells were unchanged by exposure to 10 μM cisplatin. These findings support the potential of NMR detectable UFAs to serve as a non-invasive marker of tumour cell response to treatment.

Model-based auditability of clinical trial recruitment

Detailed insight into the recruitment parameters of a clinical trial is crucial to interpretation of its results, and reasons for its success or failure. Such recruitment is increasingly done through specialized software tools, sometimes linked to Electronic Health Record (EHR) systems, enabling automated capture of audit logs. However, in the absence of shared semantic models underpinning these logs, gathered data remains insular and opaque. We propose a standardized syntactical representation to capture the provenance of the recruitment task, and ground it in CRIM, a variant of the established PCROM information model for research in primary care. The method has been successfully prototyped in the EU FP7 TRANSFoRm project, where the recruitment eligibility query module has been integrated with a provenance capture infrastructure, resulting in the full reproducibility of the study design process.

Reuse of routinely collected health data for clinical research: considerations in a central England case study

An increasing amount of electronic health data is being collected for patient care, and with it the opportunity to explore reusing this data, for example to support clinical research. We explore how researchers can be supported in identifying potential study participants using a semi-automated approach to query anonymised datasets remotely. This paper describes the socio-technical considerations when answering this question in a central England case study.

MRI texture analysis in paediatric oncology: a preliminary study

Despite current advances in neuro-imaging, the characterisation of paediatric brain tumours and neurological disorders is very challenging. Whilst Magnetic Resonance Imaging (MRI) provides images of superb clarity, it gives little information on how aggressive a tumour is. It is also very difficult to visually inspect any underlying textural patterns between tumours on MR images. This gives rise to the need for a quantitative means of analysing MR images for characterising tumours. In this work, we present a preliminary investigation into the effectiveness of texture analysis as a quantitative approach for classifying paediatric brain tumours.

A model for the electronic support of practice-based research networks

PURPOSE

The principal goal of the electronic Primary Care Research Network (ePCRN) is to enable the development of an electronic infrastructure to support clinical research activities in primary care practice-based research networks (PBRNs). We describe the model that the ePCRN developed to enhance the growth and to expand the reach of PBRN research.

METHODS

Use cases and activity diagrams were developed from interviews with key informants from 11 PBRNs from the United States and United Kingdom. Discrete functions were identified and aggregated into logical components. Interaction diagrams were created, and an overall composite diagram was constructed describing the proposed software behavior. Software for each component was written and aggregated, and the resulting prototype application was pilot tested for feasibility. A practical model was then created by separating application activities into distinct software packages based on existing PBRN business rules, hardware requirements, network requirements, and security concerns.

RESULTS

We present an information architecture that provides for essential interactions, activities, data flows, and structural elements necessary for providing support for PBRN translational research activities. The model describes research information exchange between investigators and clusters of independent data sites supported by a contracted research director. The model was designed to support recruitment for clinical trials, collection of aggregated anonymous data, and retrieval of identifiable data from previously consented patients across hundreds of practices.

CONCLUSIONS

The proposed model advances our understanding of the fundamental roles and activities of PBRNs and defines the information exchange commonly used by PBRNs to successfully engage community health care clinicians in translational research activities. By describing the network architecture in a language familiar to that used by software developers, the model provides an important foundation for the development of electronic support for essential PBRN research activities.

How to confidently teach EBM on foot: development and evaluation of a web-based e-learning course

BACKGROUND

Scarcity of well-trained clinical tutors is a key constraint in integrating teaching of evidence-based medicine (EBM) into clinical activities.

OBJECTIVES

We developed a web-based educational course for clinical trainers to confidently teach EBM principles in everyday practice. Its e-learning modules defined the learning objectives and incorporated video clips of practical and effective EBM teaching methods for exploiting educational opportunities in six different clinical settings.

METHODS

We evaluated the course with clinical tutors in different specialties across six European countries using a questionnaire to capture learning achievement against preset objectives.

RESULTS

Among 56 tutors, 47 participants (84%) improved their scores from baseline. The mean pre-course score was 69.2 (SD=10.4), which increased to 77.3 (SD=11.7) postcourse (p<0.0001). The effect size was moderate with a Cohen's d of 0.73.

CONCLUSIONS

An e-learning approach incorporating videos of applied EBM teaching and learning based on real clinical scenarios in the workplace can be useful in facilitating EBM teaching on foot. It can be integrated in the continuing professional development programmes for clinical trainers.

Dissociating effects of stimulus identity and load on working memory attentional guidance: Lengthening encoding time eliminates the effect of load but not identity

Effects of the identity and load of items in working memory (WM) on visual attention were examined. With a short interval between the WM item and a subsequent search task, there were effects of both load (slowed overall reaction times, RTs, in a WM condition relative to a mere repetition baseline) and identity (search RTs were affected by re-presentation of the item in WM in the search display). As the time to encode the initial display increased, the effects of load decreased while the effect of identity remained. The data indicate that the identity of stimuli in WM can affect the subsequent deployment of attention even when time is allowed for consolidation of the stimuli in WM, and that the WM effects are not causally related to the presence of cognitive load. The results are consistent with the identity of stimuli in WM modulating attention post the memory consolidation stage.

The lipid composition of isolated cytoplasmic lipid droplets from a human cancer cell line, BE(2)M17

(1)H nuclear magnetic resonance spectroscopy (NMR) resonances from lipids in tumours are associated with tumour grade and treatment response. The origin of these NMR signals is mainly considered to be cytoplasmic lipid droplets (LDs). Techniques exist for isolating LDs but little is known about their composition and its relationship to NMR signals. In this work, density-gradient ultracentrifugation was performed on homogenised human cancer cells to isolate LDs. (1)H NMR was performed on whole cells, isolated LDs and their extracts. Heteronuclear single quantum coherence spectroscopy (HSQC) and liquid chromatography mass spectroscopy (LC-MS) were performed on lipid extracts of LDs. Staining and microscopy were used to characterize isolated LDs. An excellent agreement in chemical shift and relative signal intensity was observed between lipid resonances in cells and isolated LD spectra supporting that NMR-visible lipids originate primarily from LDs. Isolated LDs showed high concentrations of unsaturated lipids, a oleic-to-linoleic acid ratio greater than two and a cholesteryl ester (ChE)-to-cholesterol (Ch) ratio close to unity. These ratios were several-fold greater than respective ratios in whole cells, demonstrating isolation is important to characterize LD composition. LDs contain a specific group of lipid species that are likely to contribute to the (1)H NMR spectrum of cells.

A hybrid method of application of independent component analysis to in vivo 1H MR spectra of childhood brain tumours

Independent component analysis (ICA) can automatically extract individual metabolite, macromolecular and lipid (MMLip) components from a series of in vivo MR spectra. The traditional feature extraction (FE)-based ICA approach is limited, in that a large sample size is required and a combination of metabolite and MMLip components can appear in the same independent component. The alternative ICA approach, based on blind source separation (BSS), is weak when dealing with overlapping peaks. Combining the advantages of both BSS and FE methods may lead to better results. Thus, we propose an ICA approach involving a hybrid of the BSS and FE techniques for the automated decomposition of a series of MR spectra. Experiments were performed on synthesised and patient in vivo childhood brain tumour MR spectra datasets. The hybrid ICA method showed an improvement in the decomposition ability compared with BSS-ICA or FE-ICA, with an increased correlation between the independent components and simulated metabolite and MMLip signals. Furthermore, we were able to automatically extract metabolites from the patient MR spectra dataset that were not in commonly used basis sets (e.g. guanidinoacetate).

Cohort identification for clinical research: querying federated electronic healthcare records using controlled vocabularies and semantic types

In the United Kingdom (UK), local initiatives have started to federate electronic healthcare records from different primary care clinical systems, mainly for the purposes of ensuring that health care services effectively meet the needs of the population. The use of such information is being investigated for clinical research, notably in patient cohort identification and recruitment. To achieve these aims, it is essential that the information from different systems can be searched from a single interface. While interoperability is a widely researched topic, interoperable methods and data sources in primary care are largely missing. This paper describes our approach to enabling primary care data in England to be searchable on a platform developed for performing large national collaborative primary care research studies throughout the United States.

Envisioning a learning health care system: the electronic primary care research network, a case study

PURPOSE

The learning health care system refers to the cycle of turning health care data into knowledge, translating that knowledge into practice, and creating new data by means of advanced information technology. The electronic Primary Care Research Network (ePCRN) was a project, funded by the U.S. National Institutes of Health, with the aim to facilitate clinical research using primary care electronic health records (EHRs).

METHODS

We identified the requirements necessary to deliver clinical studies via a distributed electronic network linked to EHRs. After we explored a variety of informatics solutions, we constructed a functional prototype of the software. We then explored the barriers to adoption of the prototype software within U.S. practice-based research networks.

RESULTS

We developed a system to assist in the identification of eligible cohorts from EHR data. To preserve privacy, counts and flagging were performed remotely, and no data were transferred out of the EHR. A lack of batch export facilities from EHR systems and ambiguities in the coding of clinical data, such as blood pressure, have so far prevented a full-scale deployment. We created an international consortium and a model for sharing further ePCRN development across a variety of ongoing projects in the United States and Europe.

CONCLUSIONS

A means of accessing health care data for research is not sufficient in itself to deliver a learning health care system. EHR systems need to use sophisticated tools to capture and preserve rich clinical context in coded data, and business models need to be developed that incentivize all stakeholders from clinicians to vendors to participate in the system.

Standardised representation of healthcare information in secondary care research: a Central England case study

Interoperability and reuse of healthcare information, for patient care and clinical research, rely on standardised approaches in its representation. The medical domain is rich with terminologies and dictionaries to support the representation of clinical data, but these are not necessarily interoperable. Secondary care research has long been conducted with study-specific requirements, and the valuable information collected was, however, difficult to reuse and share due to incomplete data collection and its non-standard representation. As a way to resolve some of these issues, we are designing and building sustainable database applications that clinicians and researchers alike can use as research registries, with the main aim to have research-quality healthcare information for future research studies and trials. This paper looks at a case study of how these applications are being developed in Central England through the use of controlled vocabularies. Specialty teams are keen to improve the interoperability, sharing and reuse of health information within and across specialties.

A loosely coupled framework for terminology controlled distributed EHR search for patient cohort identification in clinical research

Heterogeneous data models and coding schemes for electronic health records present challenges for automated search across distributed data sources. This paper describes a loosely coupled software framework based on the terminology controlled approach to enable the interoperation between the search interface and heterogeneous data sources. Software components interoperate via common terminology service and abstract criteria model so as to promote component reuse and incremental system evolution.

Short echo time single voxel 1H magnetic resonance spectroscopy in the diagnosis and characterisation of pineal tumours in children

BACKGROUND

Magnetic resonance spectroscopy (MRS) has been successful in characterising a range of brain tumours and is a useful aid to non-invasive diagnosis. The pineal region poses considerable surgical challenges and a major surgical resection is not required in the management of all tumours. Improved non-invasive assessment of pineal region tumours would be of considerable benefit.

METHODS

Single voxel MRS (TE 30 ms, TR 1500, 1.5 T) was performed on 15 pineal tumours: 5 germinomas, 1 non-germinomatous secreting germ cell tumour (GCT), 2 teratomas, 5 pineoblastomas, 1 pineal parenchymal tumour (PPT) of intermediate differentiation and 1 pineocytoma. Two germinomas outside the pineal gland were also studied. Metabolite, lipid and macromolecule concentrations were determined with LCModel™.

RESULTS

Germ cell tumours had significantly higher lipid and macromolecule concentrations than other tumours (t-test; P < 0.05). The teratomas had significantly lower total choline and creatine levels than germinomas (z test; P < 0.05). Taurine was convincingly detected in germinomas as well as PPTs.

CONCLUSIONS

Magnetic resonance spectroscopy is useful for characterising pineal region tumours, aiding the non-invasive diagnosis and giving additional biological insight.

In vitro metabonomic study detects increases in UDP-GlcNAc and UDP-GalNAc, as early phase markers of cisplatin treatment response in brain tumor cells

O-linked β-N-acetylglucosamine glycosylation (O-GlcNAcylation) is important in a number of biological processes and diseases including transcription, cell stress, diabetes, and neurodegeneration and may be a marker of tumor metastasis. Uridine diphospho-N-acetylglucosamine (UDP-GlcNAc), the donor molecule in O-GlcNAcylation, can be detected by (1)H nuclear magnetic resonance spectroscopy ((1)H NMR), giving the potential to measure its level noninvasively, providing a novel biomarker of prognosis and treatment monitoring. In this in vitro metabonomic study, four brain cancer cell lines were exposed to cisplatin and studied for metabolic responses using (1)H NMR. The Alamar blue assay and DAPI staining were used to assess cell sensitivity to cisplatin treatment and to confirm cell death. It is shown that in the cisplatin responding cells, UDP-GlcNAc and uridine diphospho-N-acetylgalactosamine (UDP-GalNAc), in parallel with (1)H NMR detected lipids, increased with cisplatin exposure before or at the onset of the microscopic signs of evolving cell death. The changes in UDP-GlcNAc and UDP-GalNAc were not detected in the nonresponders. These glycosylated UDP compounds, the key substrates for glycosylation of proteins and lipids, are commonly implicated in cancer proliferation and malignant transformation. However, the present study mechanistically links UDP-GlcNAc and UDP-GalNAc to cancer cell death following chemotherapeutic treatment.

A constrained least-squares approach to the automated quantitation of in vivo ¹H magnetic resonance spectroscopy data

Totally Automatic Robust Quantitation in NMR (TARQUIN), a new method for the fully automatic analysis of short echo time in vivo (1)H Magnetic resonance spectroscopy is presented. Analysis is performed in the time domain using non-negative least squares, and a new method for applying soft constraints to signal amplitudes is used to improve fitting stability. Initial point truncation and Hankel singular value decomposition water removal are used to reduce baseline interference. Three methods were used to test performance. First, metabolite concentrations from six healthy volunteers at 3 T were compared with LCModel™. Second, a Monte-Carlo simulation was performed and results were compared with LCModel™ to test the accuracy of the new method. Finally, the new algorithm was applied to 1956 spectra, acquired clinically at 1.5 T, to test robustness to noisy, abnormal, artifactual, and poorly shimmed spectra. Discrepancies of less than approximately 20% were found between the main metabolite concentrations determined by TARQUIN and LCModel™ from healthy volunteer data. The Monte-Carlo simulation revealed that errors in metabolite concentration estimates were comparable with LCModel™. TARQUIN analyses were also found to be robust to clinical data of variable quality. In conclusion, TARQUIN has been shown to be an accurate and robust algorithm for the analysis of magnetic resonance spectroscopy data making it suitable for use in a clinical setting.

How can we teach EBM in clinical practice? An analysis of barriers to implementation of on-the-job EBM teaching and learning

INTRODUCTION

Evidence-based medicine (EBM) improves the quality of health care. Courses on how to teach EBM in practice are available, but knowledge does not automatically imply its application in teaching. We aimed to identify and compare barriers and facilitators for teaching EBM in clinical practice in various European countries.

METHODS

A questionnaire was constructed listing potential barriers and facilitators for EBM teaching in clinical practice. Answers were reported on a 7-point Likert scale ranging from not at all being a barrier to being an insurmountable barrier.

RESULTS

The questionnaire was completed by 120 clinical EBM teachers from 11 countries. Lack of time was the strongest barrier for teaching EBM in practice (median 5). Moderate barriers were the lack of requirements for EBM skills and a pyramid hierarchy in health care management structure (median 4). In Germany, Hungary and Poland, reading and understanding articles in English was a higher barrier than in the other countries.

CONCLUSION

Incorporation of teaching EBM in practice faces several barriers to implementation. Teaching EBM in clinical settings is most successful where EBM principles are culturally embedded and form part and parcel of everyday clinical decisions and medical practice.

A comparison between simulated and experimental basis sets for assessing short-TE in vivo ¹H MRS data at 1.5 T

A number of algorithms designed to determine metabolite concentrations from in vivo (1)H MRS require a collection of single metabolite spectra, known as a basis set, which can be obtained experimentally or by simulation. It has been assumed that basis sets can be used interchangeably, but no systematic study has investigated the effects of small variations in basis functions on the metabolite values obtained. The aim of this study was to compare the results of simulated with experimental basis sets when used to fit short-TE (1)H MRS data of variable quality at 1.5 T. Two hundred and twelve paediatric brain tumour spectra were included in the analysis, and each was analysed twice with LCModel™ using a simulated and experimental basis set. To determine the influence of data quality on quantification, each spectrum was assessed and 152 were classified as being of 'good' quality. Bland-Altman statistics were used to measure the agreement between the two basis sets for all available spectra and only 'good'-quality spectra. Monte-Carlo simulations were performed to investigate the influence of minor shifts in metabolite frequencies on metabolite concentration estimates. All metabolites showed good agreement between the two basis sets, and the average metabolite limits of agreement were approximately ±3.84 mM for all available data and ±0.99 mM for good-quality data. Errors obtained from the Monte-Carlo analysis were found to be more accurate than the Cramer-Rao lower bounds (CRLB) for 12 of 15 metabolites when metabolite frequency shifting was considered. For the majority of purposes, a level of agreement of ±0.99 mM between simulated and experimental basis sets is sufficiently small for them to be used interchangeably. Multiple analyses using slightly modified basis sets may be useful in estimating fitting errors, which are not predicted by CRLBs.

How are "teaching the teachers" courses in evidence based medicine evaluated? A systematic review

BACKGROUND

Teaching of evidence-based medicine (EBM) has become widespread in medical education. Teaching the teachers (TTT) courses address the increased teaching demand and the need to improve effectiveness of EBM teaching. We conducted a systematic review of assessment tools for EBM TTT courses. To summarise and appraise existing assessment methods for teaching the teachers courses in EBM by a systematic review.

METHODS

We searched PubMed, BioMed, EmBase, Cochrane and Eric databases without language restrictions and included articles that assessed its participants. Study selection and data extraction were conducted independently by two reviewers.

RESULTS

Of 1230 potentially relevant studies, five papers met the selection criteria. There were no specific assessment tools for evaluating effectiveness of EBM TTT courses. Some of the material available might be useful in initiating the development of such an assessment tool.

CONCLUSION

There is a need for the development of educationally sound assessment tools for teaching the teachers courses in EBM, without which it would be impossible to ascertain if such courses have the desired effect.

Effectiveness of an e-learning course in evidence-based medicine for foundation (internship) training

AIM

To evaluate the educational effectiveness of a clinically integrated e-learning course for teaching basic evidence-based medicine (EBM) among postgraduate medical trainees compared to a traditional lecture-based course of equivalent content.

METHODS

We conducted a cluster randomized controlled trial to compare a clinically integrated e-learning EBM course (intervention) to a lecture-based course (control) among postgraduate trainees at foundation or internship level in seven teaching hospitals in the UK West Midlands region. Knowledge gain among participants was measured with a validated instrument using multiple choice questions. Change in knowledge was compared between groups taking into account the cluster design and adjusted for covariates at baseline using generalized estimating equations (GEE) model.

RESULTS

There were seven clusters involving teaching of 237 trainees (122 in the intervention and 115 in the control group). The total number of postgraduate trainees who completed the course was 88 in the intervention group and 72 in the control group. After adjusting for baseline knowledge, there was no difference in the amount of improvement in knowledge of EBM between the two groups. The adjusted post course difference between the intervention group and the control group was only 0.1 scoring points (95% CI -1.2-1.4).

CONCLUSION

An e-learning course in EBM was as effective in improving knowledge as a standard lecture-based course. The benefits of an e-learning approach need to be considered when planning EBM curricula as it allows standardization of teaching materials and is a potential cost-effective alternative to standard lecture-based teaching.

Linked metabolites: a tool for the construction of directed metabolic graphs

Metabolic pathway diagrams provide a wealth of information on how reactions combine to perform biological functions. While pathway diagrams are arranged in a way that allows a specific area of metabolism to be visualised, the inherent complexity of each pathway makes it difficult to identify the sets of reactions linking groups of compounds; a common task for researchers attempting to explain observed correlations or looking for further compounds of potential interest to use in validation work. Here we introduce Linked Metabolites, a tool that identifies sets of reactions linking groups of compounds in the context of the KEGG pathway diagrams.