Biomedical Event Extraction with Machine Learning

Biomedical natural language processing (BioNLP) is a subfield of natural language processing, an area of computational linguistics concerned with developing programs that work with natural language: written texts and speech. Biomedical relation extraction concerns the detection of semantic relations such as protein-protein interactions (PPI) from scientific texts. The aim is to enhance information retrieval by detecting relations between concepts, not just individual concepts as with a keyword search. In recent years, events have been proposed as a more detailed alternative for simple pairwise PPI relations. Events provide a systematic, structural representation for annotating the content of natural language texts. Events are characterized by annotated trigger words, directed and typed arguments and the ability to nest other events. For example, the sentence “Protein A causes protein B to bind protein C” can be annotated with the nested event structure CAUSE(A, BIND(B, C)). Converted to such formal representations, the information of natural language texts can be used by computational applications. Biomedical event annotations were introduced by the BioInfer and GENIA corpora, and event extraction was popularized by the BioNLP'09 Shared Task on Event Extraction. In this thesis we present a method for automated event extraction, implemented as the Turku Event Extraction System (TEES). A unified graph format is defined for representing event annotations and the problem of extracting complex event structures is decomposed into a number of independent classification tasks. These classification tasks are solved using SVM and RLS classifiers, utilizing rich feature representations built from full dependency parsing. Building on earlier work on pairwise relation extraction and using a generalized graph representation, the resulting TEES system is capable of detecting binary relations as well as complex event structures. We show that this event extraction system has good performance, reaching the first place in the BioNLP'09 Shared Task on Event Extraction. Subsequently, TEES has achieved several first ranks in the BioNLP'11 and BioNLP'13 Shared Tasks, as well as shown competitive performance in the binary relation Drug-Drug Interaction Extraction 2011 and 2013 shared tasks. The Turku Event Extraction System is published as a freely available open-source project, documenting the research in detail as well as making the method available for practical applications. In particular, in this thesis we describe the application of the event extraction method to PubMed-scale text mining, showing how the developed approach not only shows good performance, but is generalizable and applicable to large-scale real-world text mining projects. Finally, we discuss related literature, summarize the contributions of the work and present some thoughts on future directions for biomedical event extraction. This thesis includes and builds on six original research publications. The first of these introduces the analysis of dependency parses that leads to development of TEES. The entries in the three BioNLP Shared Tasks, as well as in the DDIExtraction 2011 task are covered in four publications, and the sixth one demonstrates the application of the system to PubMed-scale text mining.

Spatial variation in the structure and floristic composition of "restinga" vegetation in southeastern Brazil

We examined large-scale spatial variation of structural parameters and floristic composition in open Clusia scrub, a vegetation type of the Brazilian "restingas" (sandy coastal plain vegetation). This vegetation is organized in islands separated by sandy stretches with sparse herbaceous vegetation. We located 12 sample areas on three consecutive beach ridges, lying parallel to the coastline and at different distances from the ocean, in close proximity to two lagoons (Cabiúnas and Comprida). Each sample area was divided into three strips. We used the line intercept method to sample all woody plants ³ 50 cm tall. We used nested ANOVA to verify structural variation between different sampling scales. TWINSPAN analysis was performed to examine the variation in floristic composition between areas. The overall diversity index was 3.07. Six species are repeatedly dominant throughout the entire sampling area. There was homogeneity in relation to diversity and species richness between beach ridges but not within beach ridges. Floristic composition and structural parameters did not vary in relation to distance from the sea but floristic composition did vary as a function of proximity to Cabiúnas or Comprida lagoon. Differences in plant cover between sample areas may be related to the paleoformation of this sandy coastal plain.

Multiple dormancy and maternal effect on Miconia ferruginata (Melastomataceae) seed germination, Serra de Caldas Novas, Goiás, Brazil

Species with mostly asexual reproduction are interesting subjects for germination studies since variation would be more easily linked to environmental factors. Miconia ferruginata DC. is an apomictic treelet in Brazilian cerrado areas on rocky outcrops. Germination of seeds collected from individuals occurring in the Serra de Caldas Novas State Park, Goiás, was studied in three experiments under controlled conditions. Germination characteristics differed among individuals and were correlated with altitude and soil Al content. Seeds from plants growing at lower altitudes, with lower soil aluminium content, presented malformed seeds with absence of embryo which rendered lower, but better synchronized germination. The nested analysis showed that from the total variance, 78.14% for germinability, 54.56% for uncertainty of the germination process, and 68.30% for the quantity of seeds without embryo was attributed to the altitudinal effect. Individuals nested within altitude contributed up to 16.93% for the total variance. It means that there is low variability among individuals of the same altitude and high variability among individuals from different points of the slope, making clear that for the studied population the environmental effect is stronger than the genetic component to determine the seed quality. The testa of the seeds provides a mechanical dormancy which seems to be associated also with phenolic compounds, which help to disperse germination through time. Photoblastism was also registered for seeds of this species.

Prevalence and risk factors for HBV, HCV and HDV infections among injecting drug users from Rio de Janeiro, Brazil

Viral hepatitis constitutes a major health issue, with high prevalence among injecting drug users (IDUs). The present study assessed the prevalence and risk determinants for hepatitis B, C and D viruses (HBV, HCV and HDV) infections among 102 IDUs from Rio de Janeiro, Brazil. Serological markers and HCV-RNA were detected by enzyme immunoassay and nested PCR, respectively. HCV genotyping was determined by restriction fragment length polymorphism analysis (RFLP). HBsAg, anti-HBc and anti-HBs were found in 7.8, 55.8 and 24.7% of IDUs, respectively. In the final logistic regression, HBV infection was independently associated with male homosexual intercourse within the last 5 years (odds ratio (OR) 3.1; 95% confidence interval (CI) 1.1-8.8). No subject presented anti-delta (anti-HD). Anti-HCV was detected in 69.6% of subjects, and was found to be independently associated with needle sharing in the last 6 months (OR 3.4; 95% CI 1.3-9.2) and with longer duration of iv drug use (OR 3.1; 95% CI 1.1-8.7). These data demonstrate that this population is at high risk for both HBV and HCV infection. Among IDUs from Rio de Janeiro, unprotected sexual intercourse seems to be more closely associated with HBV infection, whereas HCV is positively correlated with high risk injecting behavior. Comprehensive public health interventions targeting this population and their sexual partners must be encouraged.

alpha-Globin genes: thalassemic and structural alterations in a Brazilian population

Seven unrelated patients with hemoglobin (Hb) H disease and 27 individuals with alpha-chain structural alterations were studied to identify the alpha-globin gene mutations present in the population of Southeast Brazil. The -alpha3.7, --MED and -(alpha)20.5 deletions were investigated by PCR, whereas non-deletional alpha-thalassemia (alphaHphalpha, alphaNcoIalpha, aaNcoI, alphaIcalpha and alphaTSaudialpha) was screened with restriction enzymes and by nested PCR. Structural alterations were identified by direct DNA sequencing. Of the seven patients with Hb H disease, all of Italian descent, two had the -(alpha)20.5/-alpha3.7 genotype, one had the --MED/-alpha3.7 genotype, one had the --MED/alphaHphalpha genotype and three showed interaction of the -alpha3.7 deletion with an unusual, unidentified form of non-deletional alpha-thalassemia [-alpha3.7/(aa)T]. Among the 27 patients with structural alterations, 15 (of Italian descent) had Hb Hasharon (alpha47Asp->His) associated with the -alpha3.7 deletion, 4 (of Italian descent) were heterozygous for Hb J-Rovigo (alpha53Ala->Asp), 4 (3 Blacks and 1 Caucasian) were heterozygous for Hb Stanleyville-II (alpha78Asn->Lys) associated with the alpha+-thalassemia, 1 (Black) was heterozygous for Hb G-Pest (alpha74Asp->Asn), 1 (Caucasian) was heterozygous for Hb Kurosaki (alpha7Lys->Glu), 1 (Caucasian) was heterozygous for Hb Westmead (alpha122His->Gln), and 1 (Caucasian) was the carrier of a novel silent variant (Hb Campinas, alpha26Ala->Val). Most of the mutations found reflected the Mediterranean and African origins of the population. Hbs G-Pest and Kurosaki, very rare, and Hb Westmead, common in southern China, were initially described in individuals of ethnic origin differing from those of the carriers reported in the present study and are the first cases to be reported in the Brazilian population.

Hepatitis C virus infection among Brazilian hemophiliacs: a virological, clinical and epidemiological study

We determined and analyzed risk factors of hepatitis C virus (HCV)-infected Brazilian hemophiliacs according to their virological, clinical and epidemiological characteristics. A cross-sectional and retrospective study of 469 hemophiliacs was carried out at a Brazilian blood center starting in October 1997. The prevalence of HCV infection, HCV genotypes and factors associated with HCV RNA detection was determined. The seroprevalence of anti-HCV antibodies (ELISA-3.0) was 44.6% (209/469). Virological, clinical and epidemiological assessments were completed for 162 positive patients. There were seven (4.3%) anti-HCV seroconversions between October 1992 and October 1997. During the same period, 40.8% of the positive anti-HCV hemophiliacs had abnormal alanine transaminase (ALT) levels. Plasma HCV RNA was detected by nested-RT-PCR in 116 patients (71.6%). RFLP analysis showed the following genotype distribution: HCV-1 in 98 hemophiliacs (84.5%), HCV-3 in ten (8.6%), HCV-4 in three (2.6%), HCV-2 in one (0.9%), and not typeable in four cases (3.4%). Univariate analysis indicated that older age (P = 0.017) and abnormal ALT levels (P = 0.010) were associated with HCV viremia, while the presence of inhibitor antibodies (P = 0.024) and HBsAg (P = 0.007) represented a protective factor against the presence of HCV RNA. These findings may contribute to a better understanding of the relationship between HCV infection and hemophilia.

Characterization of bovine respiratory syncytial virus isolated in Brazil

This paper presents the first isolation of bovine respiratory syncytial virus in Brazil and its physicochemical, morphological and molecular characterization. The virus was isolated from 33 samples of nasotracheal secretions, successively inoculated into a Madin-Darby bovine kidney cell culture, which was characterized by physicochemical tests and morphological observation by electron microscopy. The Brazilian sample is an RNA pleomorphic, enveloped, thermolabile and non-hemagglutinating spicular virus. Reverse transcription, followed by nested polymerase chain reaction (nRT-PCR) assay was carried out using oligonucleotides B1, B2A, B3 and B4 for the fusion proteins (F) and B5A, B6A, B7A and B8 for the attachment protein (G). The nRT-PCR-F amplified a fragment of 481 bp corresponding to part of the gene that codes for protein F, whereas nRT-PCR-G amplified a fragment of 371 bp, in agreement with part of the G gene. The virus isolated from Brazilian samples in this study corresponded to the bovine respiratory syncytial virus, and RT-PCR proved to be useful for the diagnosis of bovine clinical samples.

Absence of hepatitis B virus DNA in patients with hepatitis C and non-A-E hepatitis in the State of São Paulo, Brazil

Occult hepatitis B virus (HBV) infection has been reported as cases in which HBV DNA was detected despite the absence of any HBV serological markers or in cases in which anti-HBc antibody was the sole marker. The aim of the present study was to determine, using the polymerase chain reaction (PCR), whether HBV infection occurs in hepatitis C and non-A-E hepatitis patients without serological evidence of hepatitis B infection in São Paulo State. Two different populations were analyzed: 1) non-A-E hepatitis patients, including 12 patients with acute and 50 patients with chronic hepatic disorders without serological evidence of infection with known hepatitis viruses; 2) 43 patients previously diagnosed as hepatitis C with positive results for anti-HCV and HCV RNA. Among hepatitis C patients, anti-HBc was detected in 18.6% of the subjects. Three different sets of primers were employed for HBV DNA detection by nested PCR, covering different HBV genes: C, S and X. HBV-DNA was not detected in any sample, whereas the positive controls did produce signals. The lack of HBV DNA detection with these pairs of primers could be due to a very low viral load or to the presence of mutations in their annealing sites. The latter is unlikely as these primers were screened against an extensive dataset of HBV sequences. The development of more sensitive methods, such as real time PCR, to detect circular covalent closed DNA is necessary in order to evaluate this question since previous studies have shown that cryptic hepatitis B might occur.

Geographic distribution of hepatitis C virus genotypes in Brazil

Brazil is a country of continental dimension with a population of different ethnic backgrounds. Thus, a wide variation in the frequencies of hepatitis C virus (HCV) genotypes is expected to occur. To address this point, 1,688 sequential samples from chronic HCV patients were analyzed. HCV-RNA was amplified by the RT-PCR from blood samples collected from 1995 to 2000 at different laboratories located in different cities from all Brazilian States. Samples were collected in tubes containing a gel separator, centrifuged in the site of collection and sent by express mail in a refrigerated container to Laboratório Bioquímico Jardim Paulista, São Paulo, SP, Brazil. HCV- RNA was extracted from serum and submitted to RT and nested PCR using standard procedures. Nested PCR products were submitted to cycle sequencing reactions without prior purification. Sequences were analyzed for genotype determination and the following frequencies were found: 64.9% (1,095) for genotype 1, 4.6% (78) for genotype 2, 30.2% (510) for genotype 3, 0.2% (3) for genotype 4, and 0.1% (2) for genotype 5. The frequencies of HCV genotypes were statistically different among Brazilian regions (P = 0.00017). In all regions, genotype 1 was the most frequent (51.7 to 74.1%), reaching the highest value in the North; genotype 2 was more prevalent in the Center-West region (11.4%), especially in Mato Grosso State (25.8%), while genotype 3 was more common in the South (43.2%). Genotypes 4 and 5 were rarely found and only in the Southeast, in São Paulo State. The present data indicate the need for careful epidemiological surveys throughout Brazil since knowing the frequency and distribution of the genotypes would provide key information for understanding the spread of HCV.

HEV, TTV and GBV-C/HGV markers in patients with acute viral hepatitis

The aim of the present study was to evaluate the prevalence of HEV, TTV and GBV-C/GBV-C/HGV in patients with acute viral hepatitis A, B and non-A-C. We evaluated sera of 94 patients from a sentinel program who had acute hepatitis A (N = 40), B (N = 42) and non-A-C (N = 12); 71 blood donors served as controls. IgM and anti-HEV IgG antibodies were detected by enzyme immunoassay using commercial kits. TTV and GBV-C/HGV were detected by nested PCR; genotyping was done by sequencing and phylogenetic analysis. Anti-HEV IgG was present in 38, 10 and 17% of patients with hepatitis A, B and non-A-C. Four patients with hepatitis A and 1 with non-A-C hepatitis also had anti-HEV IgM detected in serum. TTV was detected in 21% of patients with acute hepatitis and in 31% of donors. GBV-C/HGV was detected in 9% of patients with hepatitis, and in 10% of donors. We found TTV isolates of genotypes 1, 2, 3, and 4 and GBV-C/HGV isolates of genotypes 1 and 2. Mean aminotransferase levels were lower in patients who were TTV or GBV-C/HGV positive. In conclusion, the detection of anti-HEV IgM in some acute hepatitis A cases suggests co-infection with HEV and hepatitis E could be the etiology of a few cases of sporadic non-A-C hepatitis in Salvador, Brazil. TTV genotype 1, 2, 3 and 4 isolates and GBV-C/HGV genotype 1 and 2 strains are frequent in the studied population. TTV and GBV-C/HGV infection does not appear to have a role in the etiology of acute hepatitis.

Dexamethasone-induced reactivation of bovine herpesvirus type 5 latent infection in experimentally infected rabbits results in a broader distribution of latent viral DNA in the brain

Bovine herpesvirus type 5 (BHV-5) is a major agent of meningoencephalitis in cattle and establishes latent infections mainly in sensory nerve ganglia. The distribution of latent BHV-5 DNA in the brain of rabbits prior to and after virus reactivation was studied using a nested PCR. Fifteen rabbits inoculated intranasally with BHV-5 were euthanized 60 days post-inoculation (group A, N = 8) or submitted to dexamethasone treatment (2.6 mg kg-1 day-1, im, for 5 days) and euthanized 60 days later (group B, N = 7) for tissue examination. Two groups of BHV-1-infected rabbits (C, N = 3 and D, N = 3) submitted to each treatment were used as controls. Viral DNA of group A rabbits was consistently detected in trigeminal ganglia (8/8), frequently in cerebellum (5/8), anterior cerebral cortex and pons-medulla (3/8) and occasionally in dorsolateral (2/8), ventrolateral and posterior cerebral cortices, midbrain and thalamus (1/8). Viral DNA of group B rabbits showed a broader distribution, being detected at higher frequency in ventrolateral (6/7) and posterior cerebral cortices (5/7), pons-medulla (6/7), thalamus (4/7), and midbrain (3/7). In contrast, rabbits inoculated with BHV-1 harbored viral DNA almost completely restricted to trigeminal ganglia and the distribution did not change post-reactivation. These results demonstrate that latency by BHV-5 is established in several areas of the rabbit's brain and that virus reactivation leads to a broader distribution of latent viral DNA. Spread of virus from trigeminal ganglia and other areas of the brain likely contributes to this dissemination and may contribute to the recrudescence of neurological disease frequently observed upon BHV-5 reactivation.

HBV vaccination of HCV-infected patients with occult HBV infection and anti-HBc-positive blood donors

Anti-HBc positivity is a frequent cause of donation rejection at blood banks. Hepatitis B virus (HBV) infection may also occur in HBsAg-negative patients, a situation denoted occult infection. Similarly, very low levels of HBV-DNA have also been found in the sera of patients with chronic hepatitis C virus (HCV) infection, even in the absence of serum HBsAg. Initially we searched for HBV-DNA in serum of 100 blood donors and 50 HCV-infected patients who were HBsAg negative/anti-HBc positive by nested-PCR and by an HBV monitor commercial test for HBV-DNA. Anti-HBs seroconversion rates were measured in 100 blood donors and in 22 patients with chronic HCV infection after HBV vaccination to determine if the HBV vaccination could eliminate an occult HBV infection in these individuals. Occult HBV infection was detected in proportionally fewer blood donors (6/100 = 6%) than chronic hepatitis C patients (12/50 = 24%) (P < 0.05). We noted seroconversion in 6/6 (100%) HBV-DNA(+) and in 84/94 (89.4%) HBV-DNA(-) blood donors (P > 0.05). All subjects who were HBV-DNA(+) before the first dose of HBV vaccine (D1), became HBV-DNA(-) after D1, D2, and D3. Among 22 HCV-positive patients, 10 HBV-DNA(+) and 12 HBV-DNA(-), seroconversion was observed in 9/10 (90%) HBV-DNA(+) and in 9/12 (75%) HBV-DNA(-) subjects (P > 0.05). The disappearance of HBV-DNA in the majority of vaccinated patients suggests that residual HBV can be eliminated in patients with occult infection.

Co-circulation of genotypes IA and IB of hepatitis A virus in Northeast Brazil

The Northeast region is the location of most cases of acute hepatitis A virus (HAV) in Brazil. In the present study, the genotypes of HAV strains from Pernambuco State, one of most populous states in the Northeast region, were characterized. Blood samples positive for anti-HAV IgM from 145 individuals (mean age = 29.1 years), collected during 2002 and 2003, were submitted to nested RT-PCR for amplification of the 5'non-translated region (5'NTR) and VP1/2A regions of the HAV genome. The VP1/2A and 5'NTR regions were amplified in 39 and 21% of the samples, respectively. Nucleotide sequencing was carried out in 46% of VP1/2A and in 53% of 5'NTR isolates. The identity in nucleotide sequence of the VP1/2A region ranged from 93.6 to 100.0%. Phylogenetic analysis of the VP1/2A sequences showed that 65% belong to sub-genotype IA and 35% to sub-genotype IB. Co-circulation of both sub-genotypes was observed in the two years studied. Distinct clusters of highly related sequences were observed in both sub-genotypes, suggesting endemic circulation of HAV strains in this area. In the 5'NTR isolates, 92.7-99.2% identity was observed and two isolates presented one deletion at position 413. Phylogenetic analysis showed that genotype IA strains cluster in the tree in the same way as genotype IB strains, but one IIIA isolate from Spain clusters with genotype IB strains. These results do not allow us to state that 5'NTR could be used to genotype HAV sequences. This is the first report of co-circulation of sub-genotypes IA and IB in this region, providing additional information about the molecular epidemiology of HAV strains in Brazil.

Prenatal and perinatal risk factors for bipolar disorder

Bipolar disorder (BPD) is a severe mental disorder associated with considerable morbidity and mortality. Prenatal insults have been shown to be associated with later development of mental disorders and there is a growing interest in the potential role of prenatal and perinatal risk factors in the development of BPD. The aims of this thesis were to describe the overall study design of the Finnish Prenatal Study of Bipolar Disorders (FIPS-B) and demographic characteristics of the sample. Furthermore, it was aimed to examine the association of parental age, parental age difference, perinatal complications and maternal smoking during pregnancy with BPD. This thesis is based on FIPS-B, a nested case-control study using several nationwide registers. The cases included all people born in Finland between January 1st 1983 and December 31st 1998 and diagnosed with BPD according to the Finnish Hospital Discharge Register (FHDR) before December 31st 2008. Controls for this study were people who were without BPD, schizophrenia or diagnoses related to these disorders, identified from the Population Register Centre (PRC), and matched two-fold to the cases on sex, date of birth (+/- 30 days), and residence in Finland on the first day of diagnosis of the matched case. Conditional logistic regression models were used to examine the association between risk factors and BPD. This study included 1887 BPD cases and 3774 matched controls. The mean age at diagnosis was 19.3 years and females accounted for 68% of the cases. Mothers with the lowest educational level had the highest odds of having BPD in offspring. Being born in Eastern and Southern region of Finland increased the odds of having BPD later in life. A U-shaped distribution of odds ratio was observed between paternal age and BPD in the unadjusted analysis. Maternal age and parental age difference was not associated with BPD. Birth by planned caesarean section was associated with increased odd of BPD. Smoking during pregnancy was not associated with BPD in the adjusted analyses. Region of birth and maternal educational level were associated with BPD. Both young and old father’s age was associated with BPD. Most perinatal complications and maternal smoking during pregnancy were not associated with BPD. The findings of this thesis, considered together with previous literature, suggest that the pre- and perinatal risk factor profile varies among different psychiatric disorders.

Mesenchymal stem cells from patients with chronic myeloid leukemia do not express BCR-ABL and have absence of chimerism after allogeneic bone marrow transplant

Bone marrow is a heterogeneous cell population which includes hematopoietic and mesenchymal progenitor cells. Dysregulated hematopoiesis occurs in chronic myelogenous leukemia (CML), being caused at least in part by abnormalities in the hematopoietic progenitors. However, the role of mesenchymal stem cells (MSCs) in CML has not been well characterized. The objectives of the present study were to observe the biological characteristics of MSCs from CML patients and to determine if MSCs originate in part from donors in CML patients after bone marrow transplantation (BMT). We analyzed MSCs from 5 untreated patients and from 3 CML patients after sex-mismatched allogeneic BMT. Flow cytometry analysis revealed the typical MSC phenotype and in vitro assays showed ability to differentiate into adipocytes and osteoblasts. Moreover, although some RT-PCR data were contradictory, combined fluorescence in situ hybridization analysis showed that MSCs from CML patients do not express the bcr-abl gene. Regarding MSCs of donor origin, although it is possible to detect Y target sequence by nested PCR, the low frequency (0.14 and 0.34%) of XY cells in 2 MSC CML patients by fluorescence in situ hybridization analysis suggests the presence of contaminant hematopoietic cells and the absence of host-derived MSCs in CML patients. Therefore, we conclude that MSCs from CML patients express the typical MSC phenotype, can differentiate into osteogenic and adipogenic lineages and do not express the bcr-abl gene. MSCs cannot be found in recipients 12 to 20 months after BMT. The influence of MSCs on the dysregulation of hematopoiesis in CML patients deserves further investigation.