Cancer Hallmarks Analytics Tool (CHAT): a text mining approach to organize and evaluate scientific literature on cancer.

MOTIVATION: To understand the molecular mechanisms involved in cancer development, significant efforts are being invested in cancer research. This has resulted in millions of scientific articles. An efficient and thorough review of the existing literature is crucially important to drive new research. This time-demanding task can be supported by emerging computational approaches based on text mining which offer a great opportunity to organize and retrieve the desired information efficiently from sizable databases. One way to organize existing knowledge on cancer is to utilize the widely accepted framework of the Hallmarks of Cancer. These hallmarks refer to the alterations in cell behaviour that characterize the cancer cell. RESULTS: We created an extensive Hallmarks of Cancer taxonomy and developed automatic text mining methodology and a tool (CHAT) capable of retrieving and organizing millions of cancer-related references from PubMed into the taxonomy. The efficiency and accuracy of the tool was evaluated intrinsically as well as extrinsically by case studies. The correlations identified by the tool show that it offers a great potential to organize and correctly classify cancer-related literature. Furthermore, the tool can be useful, for example, in identifying hallmarks associated with extrinsic factors, biomarkers and therapeutics targets. AVAILABILITY AND IMPLEMENTATION: CHAT can be accessed at: http://chat.lionproject.net. The corpus of hallmark-annotated PubMed abstracts and the software are available at: http://chat.lionproject.net/about. CONTACT: simon.baker@cl.cam.ac.uk. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Automatic semantic classification of scientific literature according to the hallmarks of cancer.

MOTIVATION: The hallmarks of cancer have become highly influential in cancer research. They reduce the complexity of cancer into 10 principles (e.g. resisting cell death and sustaining proliferative signaling) that explain the biological capabilities acquired during the development of human tumors. Since new research depends crucially on existing knowledge, technology for semantic classification of scientific literature according to the hallmarks of cancer could greatly support literature review, knowledge discovery and applications in cancer research. RESULTS: We present the first step toward the development of such technology. We introduce a corpus of 1499 PubMed abstracts annotated according to the scientific evidence they provide for the 10 currently known hallmarks of cancer. We use this corpus to train a system that classifies PubMed literature according to the hallmarks. The system uses supervised machine learning and rich features largely based on biomedical text mining. We report good performance in both intrinsic and extrinsic evaluations, demonstrating both the accuracy of the methodology and its potential in supporting practical cancer research. We discuss how this approach could be developed and applied further in the future. AVAILABILITY AND IMPLEMENTATION: The corpus of hallmark-annotated PubMed abstracts and the software for classification are available at: http://www.cl.cam.ac.uk/∼sb895/HoC.html. CONTACT: simon.baker@cl.cam.ac.uk.

Gender differences in cancer susceptibility: role of oxidative stress.

Cancer is a leading cause of death worldwide and environmental factors, including chemicals, have been suggested as major etiological incitements. Cancer statistics indicates that men get more cancer than women. However, differences in the known risk factors including life style or occupational exposure only offer partial explanation. Using a text mining tool, we have investigated the scientific literature concerning male- and female-specific rat carcinogens that induced tumors only in one gender in NTP 2-year cancer bioassay. Our evaluation shows that oxidative stress, although frequently reported for both male- and female-specific rat carcinogens, was mentioned significantly more in literature concerning male-specific rat carcinogens. Literature analysis of testosterone and estradiol showed the same pattern. Tox21 high-throughput assay results, although showing only weak association of oxidative stress-related processes for male- and female-specific rat carcinogens, provide additional support. We also analyzed the literature concerning 26 established human carcinogens (IARC group 1). Oxidative stress was more frequently reported for the majority of these carcinogens, and the Tox21 data resembled that of male-specific rat carcinogens. Thus, our data, based on about 600000 scientific abstracts and Tox21 screening assays, suggest a link between male-specific carcinogens, testosterone and oxidative stress. This implies that a different cellular response to oxidative stress in men and women may be a critical factor in explaining the greater cancer susceptibility observed in men. Although the IARC carcinogens are classified as human carcinogens, their classification largely based on epidemiological evidence from male cohorts, which raises the question whether carcinogen classifications should be gender specific.

Active learning-based information structure analysis of full scientific articles and two applications for biomedical literature review.

MOTIVATION: Techniques that are capable of automatically analyzing the information structure of scientific articles could be highly useful for improving information access to biomedical literature. However, most existing approaches rely on supervised machine learning (ML) and substantial labeled data that are expensive to develop and apply to different sub-fields of biomedicine. Recent research shows that minimal supervision is sufficient for fairly accurate information structure analysis of biomedical abstracts. However, is it realistic for full articles given their high linguistic and informational complexity? We introduce and release a novel corpus of 50 biomedical articles annotated according to the Argumentative Zoning (AZ) scheme, and investigate active learning with one of the most widely used ML models-Support Vector Machines (SVM)-on this corpus. Additionally, we introduce two novel applications that use AZ to support real-life literature review in biomedicine via question answering and summarization. RESULTS: We show that active learning with SVM trained on 500 labeled sentences (6% of the corpus) performs surprisingly well with the accuracy of 82%, just 2% lower than fully supervised learning. In our question answering task, biomedical researchers find relevant information significantly faster from AZ-annotated than unannotated articles. In the summarization task, sentences extracted from particular zones are significantly more similar to gold standard summaries than those extracted from particular sections of full articles. These results demonstrate that active learning of full articles' information structure is indeed realistic and the accuracy is high enough to support real-life literature review in biomedicine. AVAILABILITY: The annotated corpus, our AZ classifier and the two novel applications are available at http://www.cl.cam.ac.uk/yg244/12bioinfo.html

Weakly supervised learning of information structure of scientific abstracts--is it accurate enough to benefit real-world tasks in biomedicine?

MOTIVATION: Many practical tasks in biomedicine require accessing specific types of information in scientific literature; e.g. information about the methods, results or conclusions of the study in question. Several approaches have been developed to identify such information in scientific journal articles. The best of these have yielded promising results and proved useful for biomedical text mining tasks. However, relying on fully supervised machine learning (ml) and a large body of annotated data, existing approaches are expensive to develop and port to different tasks. A potential solution to this problem is to employ weakly supervised learning instead. In this article, we investigate a weakly supervised approach to identifying information structure according to a scheme called Argumentative Zoning (az). We apply four weakly supervised classifiers to biomedical abstracts and evaluate their performance both directly and in a real-life scenario in the context of cancer risk assessment. RESULTS: Our best weakly supervised classifier (based on the combination of active learning and self-training) performs well on the task, outperforming our best supervised classifier: it yields a high accuracy of 81% when just 10% of the labeled data is used for training. When cancer risk assessors are presented with the resulting annotated abstracts, they find relevant information in them significantly faster than when presented with unannotated abstracts. These results suggest that weakly supervised learning could be used to improve the practical usefulness of information structure for real-life tasks in biomedicine.

Gender differences in chemical carcinogenesis in National Toxicology Program 2-year bioassays.

Differences in cancer incidences between men and women are often explained by either differences in environmental exposures or by influences of sex hormones. However, there are few studies on intrinsic gender differences in susceptibility to chemical carcinogens. We have analyzed the National Toxicology Program (NTP) database for sex differences in rat responses to chemical carcinogens. We found that the odds that male rat bioassays were assigned a higher level of evidence than female rat bioassays was 1.69 (p < .001). Of 278 carcinogenic chemicals in the database, 201 (72%) exhibited statistical gender differences (p ≤ .05) in at least one nonreproductive organ. One hundred thirty of these 201 chemicals induced gender-specific tumors in male rats and 59 in female rats. Sixty-eight chemicals induced tumors in males but no tumors in females. Less than one third (i.e., 19 chemicals) induced tumors in females but not males. Male-specific tumors included pancreatic and skin tumors, and female-specific tumors included lung tumors. For some tumor sites, these differences in gender susceptibility can be associated with literature data on sex hormone receptor expression. In conclusion, gender-specific tumors were common. The male dominance is in line with recent human data, and the male susceptibility to carcinogens should be further studied.

A comparison and user-based evaluation of models of textual information structure in the context of cancer risk assessment.

BACKGROUND: Many practical tasks in biomedicine require accessing specific types of information in scientific literature; e.g. information about the results or conclusions of the study in question. Several schemes have been developed to characterize such information in scientific journal articles. For example, a simple section-based scheme assigns individual sentences in abstracts under sections such as Objective, Methods, Results and Conclusions. Some schemes of textual information structure have proved useful for biomedical text mining (BIO-TM) tasks (e.g. automatic summarization). However, user-centered evaluation in the context of real-life tasks has been lacking. METHODS: We take three schemes of different type and granularity--those based on section names, Argumentative Zones (AZ) and Core Scientific Concepts (CoreSC)--and evaluate their usefulness for a real-life task which focuses on biomedical abstracts: Cancer Risk Assessment (CRA). We annotate a corpus of CRA abstracts according to each scheme, develop classifiers for automatic identification of the schemes in abstracts, and evaluate both the manual and automatic classifications directly as well as in the context of CRA. RESULTS: Our results show that for each scheme, the majority of categories appear in abstracts, although two of the schemes (AZ and CoreSC) were developed originally for full journal articles. All the schemes can be identified in abstracts relatively reliably using machine learning. Moreover, when cancer risk assessors are presented with scheme annotated abstracts, they find relevant information significantly faster than when presented with unannotated abstracts, even when the annotations are produced using an automatic classifier. Interestingly, in this user-based evaluation the coarse-grained scheme based on section names proved nearly as useful for CRA as the finest-grained CoreSC scheme. CONCLUSIONS: We have shown that existing schemes aimed at capturing information structure of scientific documents can be applied to biomedical abstracts and can be identified in them automatically with an accuracy which is high enough to benefit a real-life task in biomedicine.

The first step in the development of Text Mining technology for Cancer Risk Assessment: identifying and organizing scientific evidence in risk assessment literature.

BACKGROUND: One of the most neglected areas of biomedical Text Mining (TM) is the development of systems based on carefully assessed user needs. We have recently investigated the user needs of an important task yet to be tackled by TM -- Cancer Risk Assessment (CRA). Here we take the first step towards the development of TM technology for the task: identifying and organizing the scientific evidence required for CRA in a taxonomy which is capable of supporting extensive data gathering from biomedical literature. RESULTS: The taxonomy is based on expert annotation of 1297 abstracts downloaded from relevant PubMed journals. It classifies 1742 unique keywords found in the corpus to 48 classes which specify core evidence required for CRA. We report promising results with inter-annotator agreement tests and automatic classification of PubMed abstracts to taxonomy classes. A simple user test is also reported in a near real-world CRA scenario which demonstrates along with other evaluation that the resources we have built are well-defined, accurate, and applicable in practice. CONCLUSION: We present our annotation guidelines and a tool which we have designed for expert annotation of PubMed abstracts. A corpus annotated for keywords and document relevance is also presented, along with the taxonomy which organizes the keywords into classes defining core evidence for CRA. As demonstrated by the evaluation, the materials we have constructed provide a good basis for classification of CRA literature along multiple dimensions. They can support current manual CRA as well as facilitate the development of an approach based on TM. We discuss extending the taxonomy further via manual and machine learning approaches and the subsequent steps required to develop TM technology for the needs of CRA.

Chemical induced alterations in p53 signaling.

The p53 protein is one of the most important tumor suppressors. The present review summarizes aspects of p53 function and its role in cancer development. Some of the most well-characterized molecular mechanisms affecting p53 regulation, stabilization, inactivation and downstream events are described. A major focus is on how xenobiotics can interfere with p53 function and on its role in chemical carcinogenesis. In the final section of this chapter we discuss future aspects on how knowledge about p53 can be used in testing of carcinogens and in risk assessment.

Text mining for improved exposure assessment.

Chemical exposure assessments are based on information collected via different methods, such as biomonitoring, personal monitoring, environmental monitoring and questionnaires. The vast amount of chemical-specific exposure information available from web-based databases, such as PubMed, is undoubtedly a great asset to the scientific community. However, manual retrieval of relevant published information is an extremely time consuming task and overviewing the data is nearly impossible. Here, we present the development of an automatic classifier for chemical exposure information. First, nearly 3700 abstracts were manually annotated by an expert in exposure sciences according to a taxonomy exclusively created for exposure information. Natural Language Processing (NLP) techniques were used to extract semantic and syntactic features relevant to chemical exposure text. Using these features, we trained a supervised machine learning algorithm to automatically classify PubMed abstracts according to the exposure taxonomy. The resulting classifier demonstrates good performance in the intrinsic evaluation. We also show that the classifier improves information retrieval of chemical exposure data compared to keyword-based PubMed searches. Case studies demonstrate that the classifier can be used to assist researchers by facilitating information retrieval and classification, enabling data gap recognition and overviewing available scientific literature using chemical-specific publication profiles. Finally, we identify challenges to be addressed in future development of the system.

Dietary sphingolipids suppress a subset of preneoplastic rat liver lesions exhibiting high PTEN, low phospho-Akt and high levels of ceramide species.

Rat liver glutathione-S-transferase Pi-(GST-P)-positive enzyme-altered foci (EAF) are preneoplastic lesions that develop in response to carcinogenic stress. They are often used as endpoints in e.g. chemopreventive studies. In this study we characterize a pAkt-negative/ceramide-positive (pAkt-/cer+) EAF phenotype, as defined by immunohistochemistry for pAkt and ceramide species, in diethylnitrosamine(DEN)-, phenobarbital- or aflatoxinB1-treated rats. There was a close to 100% overlap for the pAkt and the ceramide marker. Furthermore, serial sections stained for PTEN indicated a close correlation between PTEN-positive and pAkt-negative lesions in DEN-treated rats. Experiments with DEN-treated rats given sphingomyelin in the diet suggested that sphingomyelin selectively targeted these lesions. In in vitro experiments sphingosine rapidly decreased pAkt levels in hepatocytes, and in experiments with hepatocytes from DEN-treated rats sphingosine selectively killed EAF cells. Furthermore, pretreatment with antisense Akt oligonucleotides in vitro sensitized non-EAF hepatocytes, so that EAF and non-EAF cells became equally sensitive to sphingosine. It is concluded that rat liver, in response to carcinogenic stress, develops a distinct EAF phenotype exhibiting low pAkt levels and concomitant alterations in sphingolipid metabolism. Our data also suggest that pAkt-/cer+ EAF are selectively targeted by sphingolipids in the diet and that lesions with this phenotype should be of particular interest for future studies on chemopreventive effects that may affect sphingolipid metabolism.

Combining QSAR Modeling and Text-Mining Techniques to Link Chemical Structures and Carcinogenic Modes of Action.

There is an increasing need for new reliable non-animal based methods to predict and test toxicity of chemicals. Quantitative structure-activity relationship (QSAR), a computer-based method linking chemical structures with biological activities, is used in predictive toxicology. In this study, we tested the approach to combine QSAR data with literature profiles of carcinogenic modes of action automatically generated by a text-mining tool. The aim was to generate data patterns to identify associations between chemical structures and biological mechanisms related to carcinogenesis. Using these two methods, individually and combined, we evaluated 96 rat carcinogens of the hematopoietic system, liver, lung, and skin. We found that skin and lung rat carcinogens were mainly mutagenic, while the group of carcinogens affecting the hematopoietic system and the liver also included a large proportion of non-mutagens. The automatic literature analysis showed that mutagenicity was a frequently reported endpoint in the literature of these carcinogens, however, less common endpoints such as immunosuppression and hormonal receptor-mediated effects were also found in connection with some of the carcinogens, results of potential importance for certain target organs. The combined approach, using QSAR and text-mining techniques, could be useful for identifying more detailed information on biological mechanisms and the relation with chemical structures. The method can be particularly useful in increasing the understanding of structure and activity relationships for non-mutagens.

Grouping chemicals for health risk assessment: A text mining-based case study of polychlorinated biphenyls (PCBs).

As many chemicals act as carcinogens, chemical health risk assessment is critically important. A notoriously time consuming process, risk assessment could be greatly supported by classifying chemicals with similar toxicological profiles so that they can be assessed in groups rather than individually. We have previously developed a text mining (TM)-based tool that can automatically identify the mode of action (MOA) of a carcinogen based on the scientific evidence in literature, and it can measure the MOA similarity between chemicals on the basis of their literature profiles (Korhonen et al., 2009, 2012). A new version of the tool (2.0) was recently released and here we apply this tool for the first time to investigate and identify meaningful groups of chemicals for risk assessment. We used published literature on polychlorinated biphenyls (PCBs)-persistent, widely spread toxic organic compounds comprising of 209 different congeners. Although chemically similar, these compounds are heterogeneous in terms of MOA. We show that our TM tool, when applied to 1648 PubMed abstracts, produces a MOA profile for a subgroup of dioxin-like PCBs (DL-PCBs) which differs clearly from that for the rest of PCBs. This suggests that the tool could be used to effectively identify homogenous groups of chemicals and, when integrated in real-life risk assessment, could help and significantly improve the efficiency of the process.

Evaluation of carcinogenic modes of action for pesticides in fruit on the Swedish market using a text-mining tool.

Toxicity caused by chemical mixtures has emerged as a significant challenge for toxicologists and risk assessors. Information on individual chemicals' modes of action is an important part of the hazard identification step. In this study, an automatic text mining-based tool was employed as a method to identify the carcinogenic modes of action of pesticides frequently found in fruit on the Swedish market. The current available scientific literature on the 26 most common pesticides found in apples and oranges was evaluated. The literature was classified according to a taxonomy that specifies the main type of scientific evidence used for determining carcinogenic properties of chemicals. The publication profiles of many pesticides were similar, containing evidence for both genotoxic and non-genotoxic modes of action, including effects such as oxidative stress, chromosomal changes and cell proliferation. We also found that 18 of the 26 pesticides studied here had previously caused tumors in at least one animal species, findings which support the mode of action data. This study shows how a text-mining tool could be used to identify carcinogenic modes of action for a group of chemicals in large quantities of text. This strategy could support the risk assessment process of chemical mixtures.

Data and literature gathering in chemical cancer risk assessment.

In recent years, chemical cancer risk assessment has faced major challenges: the demand for cancer risk assessment has grown considerably with strict legislation regarding chemical safety, whereas cancer hazard identification has turned increasingly complex due to the rapid development and high publication rate in biomedical sciences. Thus, much of the scientific evidence required for hazard identification is hidden in large collections of biomedical literature. Extensive guidelines have been produced to support cancer risk assessment under these circumstances. We evaluated whether these guidelines support the first, critical step of this task--data and literature gathering--and found that the guidance is vague. We propose ways to improve data and literature gathering for cancer risk assessment and suggest developing a computational literature search and analysis tool dedicated to the task. We describe the first prototype tool we have developed and discuss how it could help to improve the quality, consistency, and effectiveness of cancer risk assessment when developed further. Fully reliable automatic data and literature gathering may not be realistic; the retrieved articles will always need to be examined further by risk assessors. However, our proposal offers a starting point for improved data and literature gathering that can benefit the whole cancer risk assessment process.

Text mining for literature review and knowledge discovery in cancer risk assessment and research.

Research in biomedical text mining is starting to produce technology which can make information in biomedical literature more accessible for bio-scientists. One of the current challenges is to integrate and refine this technology to support real-life scientific tasks in biomedicine, and to evaluate its usefulness in the context of such tasks. We describe CRAB - a fully integrated text mining tool designed to support chemical health risk assessment. This task is complex and time-consuming, requiring a thorough review of existing scientific data on a particular chemical. Covering human, animal, cellular and other mechanistic data from various fields of biomedicine, this is highly varied and therefore difficult to harvest from literature databases via manual means. Our tool automates the process by extracting relevant scientific data in published literature and classifying it according to multiple qualitative dimensions. Developed in close collaboration with risk assessors, the tool allows navigating the classified dataset in various ways and sharing the data with other users. We present a direct and user-based evaluation which shows that the technology integrated in the tool is highly accurate, and report a number of case studies which demonstrate how the tool can be used to support scientific discovery in cancer risk assessment and research. Our work demonstrates the usefulness of a text mining pipeline in facilitating complex research tasks in biomedicine. We discuss further development and application of our technology to other types of chemical risk assessment in the future.

Exocrine pancreatic carcinogenesis and autotaxin expression.

Exocrine pancreatic cancer is an aggressive disease with an exceptionally high mortality rate. Genetic analysis suggests a causative role for environmental factors, but consistent epidemiological support is scarce and no biomarkers for monitoring the effects of chemical pancreatic carcinogens are available. With the objective to identify common traits for chemicals inducing pancreatic tumors we studied the National Toxicology Program (NTP) bioassay database. We found that male rats were affected more often than female rats and identified eight chemicals that induced exocrine pancreatic tumors in males only. For a hypothesis generating process we used a text mining tool to analyse published literature for suggested mode of actions (MOA). The resulting MOA analysis suggested inflammatory responses as common feature. In cell studies we found that all the chemicals increased protein levels of the inflammatory protein autotaxin (ATX) in Panc-1, MIA PaCa-2 or Capan-2 cells. Induction of MMP-9 and increased invasive migration were also frequent effects, consistent with ATX activation. Testosterone has previously been implicated in pancreatic carcinogenesis and we found that it increased ATX levels. Our data show that ATX is a target for chemicals inducing pancreatic tumors in rats. Several lines of evidence implicate ATX and its product lysophosphatidic acid in human pancreatic cancer. Mechanisms of action may include stimulated invasive growth and metastasis. ATX may interact with hormones or onco- or suppressor-genes often deregulated in exocrine pancreatic cancer. Our data suggest that ATX is a target for chemicals promoting pancreatic tumor development.

Combined toxic exposures and human health: biomarkers of exposure and effect.

Procedures for risk assessment of chemical mixtures, combined and cumulative exposures are under development, but the scientific database needs considerable expansion. In particular, there is a lack of knowledge on how to monitor effects of complex exposures, and there are few reviews on biomonitoring complex exposures. In this review we summarize articles in which biomonitoring techniques have been developed and used. Most examples describe techniques for biomonitoring effects which may detect early changes induced by many chemical stressors and which have the potential to accelerate data gathering. Some emphasis is put on endocrine disrupters acting via epigenetic mechanisms and on carcinogens. Solid evidence shows that these groups of chemicals can interact and even produce synergistic effects. They may act during sensitive time windows and biomonitoring their effects in epidemiological studies is a challenging task.

Induction of preneoplastic rat liver lesions with an attenuated p53 response by low doses of diethylnitrosamine.

Previous reports have documented an attenuated p53 response to DNA-damage in preneoplastic enzyme-altered foci (EAF). Data suggest that this alteration is an adaptation to genotoxic stress induced by carcinogens. Here, we have studied whether the altered p53 response in EAF can be related to acutely apoptotic or cytotoxic doses of the carcinogen diethylnitrosamine (DEN). Eight groups of rats received cumulative doses of 0.25, 0.5, 1.0 and 2.0 mmol DEN/kg, administered weekly for either 10 or 20 weeks. A ninth group received 3.0 mmol/kg for 10 weeks, which gave a supralinear EAF response. Twenty-four hours before sacrifice, all rats were given a challenge dose of DEN in order to induce a p53 response in hepatocytes. The numbers of p53-positive hepatocytes in EAF and in surrounding tissue were analyzed. Unexpectedly, all cumulative doses gave rise to p53-negative EAF and 20-week treatment gave larger EAF with fewer p53-positive hepatocytes than 10-week treatment. It was also observed that at the lowest doses, most EAF developed in midzonal areas. Similar results were obtained with aflatoxin B1. Single high doses of DEN induced p53 accumulation and apoptosis within 24 h, whereas lower doses did not. It is concluded that EAF with an attenuated p53 response can be induced by low doses of genotoxic compounds, not giving rise to acute apoptosis or necrosis. Instead, it is suggested that time is an important determinant for its development at low doses and that a delayed type of apoptosis might be important.

Sphingolipids suppress preneoplastic rat hepatocytes in vitro and in vivo.

Sphingolipids can modulate cell growth, differentiation and apoptosis. In the present investigation, selective death of hepatocytes localized in enzyme-altered foci (EAF hepatocytes) was shown to be induced by sphingolipids. Sphingosine (20 micro M) caused rapid cell death predominantly of EAF hepatocytes in vitro. During 4 h of such exposure, cytochrome c was released from the mitochondria into the cytoplasm and the number of cells demonstrating cleaved caspase-9 activity increased. The selective sensitivity of EAF cells to sphingolipid-induced death was attenuated by tumor necrosis factor-alpha. In previous studies we have demonstrated that EAF hepatocytes are resistant to Fas-mediated apoptosis, a resistance shown here to be reversed by low concentrations of sphingosine. Immunohistological staining revealed higher levels of glucosylated ceramide in EAF than in the surrounding tissue. Furthermore, an inhibitor of glucosylation enhanced the toxicity of ceramide towards EAF cells. TLC analysis suggested low levels of sphingosine in preneoplastic lesions. In in vivo experiments EAF-bearing rats were fed a diet supplemented with 0.1% sphingomyelin for 2 weeks. Sphingolipid feeding reduced the number of EAF and EAF area in the liver by 40-50% as compared with rats fed a control diet. These studies indicate that the turnover of sphingolipids in preneoplastic EAF hepatocytes is altered. This alteration may explain not only the increased sensitivity of EAF cells towards sphingolipid-induced cell death, but also the resistance of these hepatocytes to cell death involving sphingolipids as second messengers. Furthermore, sphingomyelin in the diet may prevent EAF development. It is suggested that the altered turnover of sphingolipids might be a target for chemoprevention of hepatocellular carcinoma.