Analizando Redes de Colaboração Científica#
Para podermos analizar redes de colaboração em publicações científicas, precisamos antes construir um banco de referências. Para isso iremos utilizar este código. Este script foi construído como uma ferramenta para a construção de um banco de referências sobre COVID19 para as análises deste capitulo. Caso o leitor deseje, pode também modificar o script para atender seus interesses pessoais de busca bibliográficas.
Começamos abaixo com uma busca de no máximo 500 artigos por palavra-chave. Esta linha de comando irá gerar um banco de dados chamado pubmed.sqlite.
!./fetch2sql.py --count 500
Show code cell output
Fetching results for covid19
Found 500 items
27%|███████████▏ | 136/500 [01:48<04:48, 1.26it/s]
Buscas no Corpus#
Este banco de dados foi configurado com indexação automática do seu conteúdo. Logo podemos escrever uma simples função para realizar as buscas:
from sqlalchemy import create_engine, text
from IPython.core.display import display, HTML
def busca(q):
engine = create_engine(f"sqlite:///pubmed.sqlite")
with engine.connect() as con:
res = con.execute(text(f"select rowid, title,highlight(article_fts, 1, '<b>', '</b>') from article_fts('{q}');")).fetchall()
del engine
return res
for pmid, title, a in busca('cov2'):
display(pmid, title, HTML(a))
36499351
'Pathological Roles of Pulmonary Cells in Acute Lung Injury: Lessons from Clinical Practice.'
Interstitial lung diseases (ILD) are relatively rare and sometimes become life threatening. In particular, rapidly progressive ILD, which frequently presents as acute lung injury (ALI) on lung histopathology, shows poor prognosis if proper and immediate treatments are not initiated. These devastating conditions include acute exacerbation of idiopathic pulmonary fibrosis (AE-IPF), clinically amyopathic dermatomyositis (CADM), epidermal growth factor receptor-tyrosine kinase inhibitor (EGFR-TKI)-induced lung injury, and severe acute respiratory syndrome coronavirus 2 (SARS-CoV2) infection named coronavirus disease 2019 (COVID-19). In this review, clinical information, physical findings, laboratory examinations, and findings on lung high-resolution computed tomography and lung histopathology are presented, focusing on majorly damaged cells in each disease. Furthermore, treatments that should be immediately initiated in clinical practice for each disease are illustrated to save patients with these diseases.
36501018
'Predictors of Sarcopenia in Outpatients with Post-Critical SARS-CoV2 Disease. Nutritional Ultrasound of Rectus Femoris Muscle, a Potential Tool.'
The loss of muscle mass in post-critical COVID-19 outpatients is difficult to assess due to the limitations of techniques and the high prevalence of obesity. Ultrasound is an emerging technique for evaluating body composition. The aim is to evaluate sarcopenia and its risk factors, determining ultrasound usefulness as a potential tool for this purpose according to established techniques, such as the bioimpedance vector analysis (BIVA), handgrip strength, and timed up-and-go test.
36503438
'Severe COVID-19-associated myocarditis with cardiogenic shock - management with assist devices - a case report & review.'
Primary viral myocarditis associated with severe acute respiratory syndrome coronavirus 2 (SARS-Cov2) infection is a rare diagnosis.
36505528
'Seroprevalence of SARS-CoV 2 antibodies & its determinants in children of 5-to-18-year age group in an urban setting, Kerala.'
There were limited data on the true burden of COVID 19 infection in children since the majority of the infections are asymptomatic or paucisymptomatic. This study aimed to measure the prevalence of SARS CoV2 antibodies in children of the 5-to-18 years age group.
36506002
'Maternal fructose boosts the effects of a Western-type diet increasing SARS-COV-2 cell entry factors in male offspring.'
Fructose-rich beverages and foods consumption correlates with the epidemic rise in cardiovascular disease, diabetes and obesity. Severity of COVID-19 has been related to these metabolic diseases. Fructose-rich foods could place people at an increased risk for severe COVID-19. We investigated whether maternal fructose intake in offspring affects hepatic and ileal gene expression of proteins that permit SARS-CoV2 entry to the cell. Carbohydrates were supplied to pregnant rats in drinking water. Adult and young male descendants subjected to water, liquid fructose alone or as a part of a Western diet, were studied. Maternal fructose reduced hepatic SARS-CoV2 entry factors expression in older offspring. On the contrary, maternal fructose boosted the Western diet-induced increase in viral entry factors expression in ileum of young descendants. Maternal fructose intake produced a fetal programming that increases hepatic viral protection and, in contrast, exacerbates fructose plus cholesterol-induced diminution in SARS-CoV2 protection in small intestine of progeny.
36507223
'Ad26.COV2.S priming provided a solid immunological base for mRNA-based COVID-19 booster vaccination.'
The emergence of novel SARS-CoV-2 variants led to the recommendation of booster vaccinations after Ad26.COV2.S priming. It was previously shown that heterologous booster vaccination induces high antibody levels, but how heterologous boosters affect other functional aspects of the immune response remained unknown. Here, we performed immunological profiling of Ad26.COV2.S-primed individuals before and after homologous or heterologous (mRNA-1273 or BNT162b2) booster. Booster vaccinations increased functional antibodies targeting ancestral SARS-CoV-2 and emerging variants. Especially heterologous booster vaccinations induced high levels of functional antibodies. In contrast, T cell responses were similar in magnitude following homologous or heterologous booster vaccination, and retained cross-reactivity towards variants. Booster vaccination led to a minimal expansion of SARS-CoV-2-specific T cell clones and no increase in breadth of the T cell repertoire. In conclusion, we show that Ad26.COV2.S priming vaccination provided a solid immunological base for heterologous boosting, increasing humoral and cellular responses targeting emerging variants of concern.
36510222
'Longitudinal transcriptional analysis of peripheral blood leukocytes in COVID-19 convalescent donors.'
SARS-CoV2 can induce a strong host immune response. Many studies have evaluated antibody response following SARS-CoV2 infections. This study investigated the immune response and T cell receptor diversity in people who had recovered from SARS-CoV2 infection (COVID-19).
36510647
"Increased incidence of Graves'' disease during the SARS-CoV2 pandemic."
COVID-19 has a wide spectrum of clinical severity and there is evidence that SARS-Cov2 affects several organs and systems. Among the organs affected since the beginning of the pandemic, the relationship between SARS-CoV-2 infection and thyroid involvement has been demonstrated. Novel and highly effective mRNA and DNA-based vaccines have been rapidly developed to decrease SARS-CoV-2 morbidity and mortality. Early after mass vaccinations, cases of thyroid dysfunction mainly including episodes of subacute thyroiditis, began to be reported like adverse effects. The objective of this study is to determine the impact of the pandemic, both due to SARS-CoV2 infections and vaccinations, on the incidence of Graves´ disease.
36511263
'Factors Associated with Vaccine Breakthrough Incidence among Health Care Workers Vaccinated with Inactivated SARS-CoV2 Vaccine (CoronaVac).'
Healthcare workers (HCWs) run a high risk of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The HCWs are prone to the SARS-CoV-2 infection in the hospital despite being fully vaccinated. The present study aimed to address the factors associated with the coronavirus disease 2019 (COVID-19) vaccine breakthrough among HCWs.
36514704
'Remdesivir-Induced Liver Injury in a Patient With Coronavirus Disease 2019 and History of Congestive Hepatopathy.'
Remdesivir is an antiviral agent used as supportive care in adults with SARS-COV2-induced pneumonia. We report a case of an 81-year-old patient who developed hepatocellular acute liver injury 48 hours after initiating remdesivir. During the investigation, other causes of hepatotoxicity were excluded. A decrease in transaminases and international normalized ratio (INR) was observed 24 hours after cessation of remdesivir. An abdominal CT demonstrated hepatic congestion, retrograde hepatic venous opacification shortly after intravenous contrast injection, and dilatation of hepatic veins and inferior vena cava. We suggest congestive hepatopathy secondary to remdesivir as a possible component of liver injury.
Redes de Citações#
Nosso código de coleta de artigos no PubMed também construiu uma tabela de citações de cada artigo baixado.
import networkx as nx
Vamos começar com um id de artigo que possua citações. Mas antes precisamos escrever uma função para buscar os vizinhos futuros, ou seja, os artigos que os citam.
def get_neighbors(pmid):
engine = create_engine('sqlite:///pubmed.sqlite',
echo=False)
sql = f"select cited_by from citations_covid19 where pmid={pmid};"
res = engine.execute(sql).fetchone()
if res is not None:
return res[0].split('|')
Na função acima, caso não existam citações registradas, a sua saída será None.
get_neighbors(36502386)
['36502408', '36496350']
Para construir a rede de citações, vamos usar um objeto fila (queue em inglês) do tipo FIFO (first in, first out). O código abaixo começa por um artigo que será a raiz da rede e o adiciona à fila. Em seguida, usando a função get_neighbors acima encontramos os artigos que o citam, criamos arestas na rede e os adicionamos à fila. O loop while repete estas operações até que a fila esteja vazia.
from queue import SimpleQueue
q = SimpleQueue()
raiz = 36508742
q.put(raiz)
árvore = nx.DiGraph()
while True:
if q.empty():
break
nó = q.get()
vizinhos = get_neighbors(nó)
if isinstance(vizinhos, list):
for v in vizinhos:
árvore.add_edge(nó, v)
q.put(int(v))
# pos = nx.nx_agraph.graphviz_layout(árvore, prog="twopi", args="")
pos = nx.kamada_kawai_layout(árvore)
nx.draw(árvore, pos, with_labels=True)
nx.draw_networkx_nodes(árvore, pos, node_size=2000, nodelist=[raiz], node_color="tab:orange");
Estabelecendo um Gerador de Artigos#
Como de costume em análise de textos precisamos estabelecer uma metodologia de acesso gradual ao corpus para evitar problemas de consumo excessivo de memória. Neste espírito vamos começar construindo um um gerador que nos entregue um artigo por vez, a partir do nosso banco de dados.
from gensim import utils
from sqlalchemy import create_engine
class StreamCorpus(object):
def __init__(self, table, limit=100):
"""
Iterador sobre o corpus no banco SQLite
:param table: nome da tabela
:param limit: limite de artigos
"""
self.engine = create_engine('sqlite:///pubmed.sqlite',
echo=False)
self.limit = limit
self.table = table
self.conn = self.engine.connect()
@property
def text_gen(self):
return (str(text[0]) for text in
self.conn.execute(f"SELECT abstract FROM {self.table} limit {self.limit}").fetchall() if
text[0].strip() != '')
def __iter__(self):
for doc in self.text_gen:
yield list(utils.simple_tokenize(doc))
self.conn.close()
O código acima retorna os resumos tokenizados das referências armazenadas no banco de dados. Vamos examinar alguns destes resumos.
SC = StreamCorpus('covid19', 1000)
i = 0
for abs in SC:
print (abs)
if i == 2:
break
i += 1
['This', 'article', 'explores', 'the', 'question', 'of', 'why', 'the', 'nine', 'pandemics', 'prior', 'to', 'COVID', 'which', 'have', 'affected', 'millions', 'of', 'people', 'since', 'the', 'second', 'half', 'of', 'the', 'th', 'century', 'were', 'not', 'recorded', 'in', 'collective', 'memory', 'despite', 'their', 'magnitude', 'and', 'extent', 'Thus', 'it', 'proposes', 'a', 'reading', 'of', 'the', 'pandemic', 'as', 'one', 'component', 'of', 'a', 'wider', 'syndemic', 'made', 'up', 'of', 'contagious', 'diseases', 'climate', 'change', 'and', 'malnutrition', 'This', 'piece', 'offers', 'a', 'narrative', 'of', 'the', 'origins', 'development', 'and', 'prospects', 'of', 'the', 'pandemic', 'within', 'the', 'dynamics', 'of', 'the', 'global', 'food', 'system', 'and', 'national', 'economic', 'and', 'political', 'systems', 'highlighting', 'components', 'and', 'connections', 'It', 'includes', 'a', 'warning', 'that', 'along', 'with', 'climate', 'change', 'and', 'malnutrition', 'undernourishment', 'obesity', 'pandemics', 'are', 'known', 'and', 'expected', 'outcomes', 'of', 'the', 'workings', 'of', 'a', 'socio', 'political', 'system', 'that', 'as', 'in', 'the', 'case', 'of', 'other', 'components', 'of', 'the', 'syndemic', 'by', 'naturalizing', 'causes', 'and', 'individualizing', 'consequences', 'conspire', 'against', 'the', 'creation', 'of', 'narratives', 'that', 'go', 'beyond', 'cosmetic', 'changes']
['The', 'decline', 'in', 'vaccination', 'coverage', 'associated', 'with', 'the', 'population', 's', 'mistrust', 'of', 'vaccines', 'has', 'been', 'defined', 'as', 'a', 'global', 'health', 'threat', 'Adopting', 'a', 'qualitative', 'approach', 'centered', 'on', 'the', 'social', 'significance', 'of', 'vaccines', 'we', 'conducted', 'semi', 'structured', 'interviews', 'with', 'mothers', 'and', 'fathers', 'in', 'the', 'City', 'of', 'Buenos', 'Aires', 'between', 'July', 'and', 'December', 'We', 'describe', 'and', 'analyze', 'their', 'knowledge', 'and', 'the', 'arguments', 'used', 'to', 'justify', 'their', 'acceptance', 'or', 'reluctance', 'to', 'vaccinate', 'their', 'children', 'in', 'the', 'context', 'of', 'COVID', 'The', 'results', 'show', 'that', 'the', 'studied', 'population', 'possesses', 'knowledge', 'about', 'vaccines', 'and', 'accepts', 'them', 'as', 'a', 'preventive', 'practice', 'Nonetheless', 'it', 'was', 'possible', 'to', 'identify', 'a', 'broad', 'spectrum', 'of', 'attitudes', 'towards', 'vaccination', 'including', 'more', 'reticent', 'positions', 'based', 'on', 'concerns', 'regarding', 'their', 'safety', 'adverse', 'effects', 'and', 'the', 'need', 'to', 'apply', 'them', 'for', 'eradicated', 'diseases', 'proposing', 'vaccination', 'as', 'an', 'optional', 'and', 'complementary', 'practice', 'to', 'other', 'care', 'practices', 'Arguments', 'are', 'organized', 'under', 'the', 'assumption', 'of', 'individual', 'and', 'moral', 'responsibility', 'for', 'health', 'care', 'an', 'expression', 'of', 'the', 'current', 'sociocultural', 'and', 'health', 'backdrop']
['This', 'article', 'summarizes', 'the', 'strategies', 'used', 'to', 'rapidly', 'develop', 'COVID', 'vaccines', 'and', 'distribute', 'them', 'globally', 'with', 'an', 'emphasis', 'on', 'vaccines', 'developed', 'in', 'western', 'nations', 'It', 'is', 'based', 'on', 'interviews', 'and', 'information', 'gathered', 'regarding', 'the', 'response', 'to', 'the', 'pandemic', 'both', 'from', 'international', 'organizations', 'and', 'official', 'documents', 'from', 'Brazil', 'Argentina', 'Colombia', 'Peru', 'and', 'Mexico', 'While', 'vaccine', 'development', 'has', 'been', 'hailed', 'as', 'successful', 'their', 'global', 'distribution', 'has', 'been', 'highly', 'unequal', 'We', 'look', 'at', 'how', 'the', 'pandemic', 'succeeded', 'in', 'mobilizing', 'large', 'quantities', 'of', 'government', 'resources', 'and', 'how', 'citizens', 'volunteered', 'their', 'bodies', 'so', 'that', 'clinical', 'trials', 'could', 'be', 'completed', 'quickly', 'However', 'patents', 'prevented', 'the', 'expansion', 'of', 'manufacturing', 'capacity', 'and', 'the', 'governments', 'of', 'a', 'few', 'wealthy', 'countries', 'prioritized', 'the', 'protection', 'and', 'in', 'some', 'cases', 'overprotection', 'of', 'their', 'citizens', 'at', 'the', 'expense', 'of', 'protecting', 'the', 'rest', 'of', 'world', 's', 'population', 'Among', 'the', 'major', 'beneficiaries', 'of', 'the', 'global', 'response', 'to', 'the', 'pandemic', 'are', 'the', 'leading', 'vaccine', 'companies', 'their', 'executives', 'and', 'investors', 'The', 'article', 'concludes', 'with', 'some', 'of', 'the', 'lessons', 'learned', 'in', 'this', 'process']
Construindo Modelos Vetoriais#
from gensim.models import Doc2Vec, FastText, KeyedVectors, Word2Vec
from gensim.models.doc2vec import TaggedLineDocument
from gensim.models.word2vec import LineSentence
from gensim.similarities.annoy import AnnoyIndexer
from gensim.utils import save_as_line_sentence
class Embeddings:
def __init__(self,table_name, filename, corpus_size):
"""
Create embeddings based on a corpus
:param filename: file name where the corpus with be temporarily dumped.
:param corpus_size: number of documents to be used to build the embeddings
"""
self.table_name = table_name
self.filename = filename
self.corpus_size = corpus_size
def dump_to_corpus_file(self, table_name):
stream_corpus = StreamCorpus(self.table_name, self.corpus_size)
save_as_line_sentence(stream_corpus, self.filename)
def gen_wor2vec(self):
"""
Build Word2Vec embedding
:return:
"""
model = Word2Vec(sentences=LineSentence(self.filename), workers=8)
model.save(f"{self.filename.split('.')[0]}.w2v")
return model
def gen_WMD(self):
model = Word2Vec
def gen_fasttext(self):
model = FastText(sentences=LineSentence(self.filename), workers=8)
model.save(f"{self.filename.split('.')[0]}.ft")
return model
def gen_doc2vec(self):
model = Doc2Vec(documents=TaggedLineDocument(self.filename), workers=8)
model.save(f"{self.filename.split('.')[0]}.d2v")
return model
Modelo Word2Vec#
# del SC
E = Embeddings('covid19','articles.txt',1000)
E.dump_to_corpus_file('articles')
mwv = E.gen_wor2vec()
mwv.wv.most_similar('SARS')
[('syndrome', 0.9543189406394958),
('respiratory', 0.9537762403488159),
('acute', 0.9496352076530457),
('infection', 0.9480603337287903),
('coronavirus', 0.9478137493133545),
('severe', 0.9424639344215393),
('disease', 0.918355405330658),
('CoV', 0.915025532245636),
('corona', 0.914029061794281),
('caused', 0.8789240717887878)]
Modelo “Fast Text”#
O modelo Fast text nos permite lidar com variações morfológicas das palavras
E = Embeddings('covid19','articles.txt',1000)
E.dump_to_corpus_file('articles')
mft = E.gen_fasttext()
mft.wv.most_similar('SARS')
[('CoV', 0.9942441582679749),
('coronaviruses', 0.9934917092323303),
('virus', 0.993304967880249),
('infect', 0.9931924343109131),
('corona', 0.9928044676780701),
('ARS', 0.9926421642303467),
('syndrome', 0.9925889372825623),
('pseudovirus', 0.9925788640975952),
('Syndrome', 0.9924823045730591),
('Coronaviridae', 0.9921442270278931)]
Modelo Doc2Vec#
Este modelo constroi uma representação vetorial dos documentos permitindo consultas de similaridade entre artigos.
E = Embeddings('covid19','articles.txt',1000)
E.dump_to_corpus_file('articles')
mdv = E.gen_doc2vec()
Podemos encontrar quais os documentos mais similar a um texto (já tokenizado):
texto = ['Spontaneous', 'adverse', 'reaction', 'reports', 'of', 'sudden', 'hearing', 'loss', 'have', 'been',
'observed', 'and', 'a', 'population', 'based', 'cohort', 'study', 'conducted', 'in', 'Israel', 'showed',
'an', 'increase', 'in', 'the', 'incidence', 'of', 'sudden', 'sensorineural', 'hearing', 'loss', 'SSNHL',
'following', 'vaccination', 'with', 'messenger', 'RNA', 'COVID', 'vaccine', 'BNT', 'b', 'Pfizer', 'BioNTech',
'However', 'in', 'this', 'setting', 'the', 'possibility', 'of', 'confounding', 'remained']
inferred_vector = mdv.infer_vector(texto)
sims = mdv.dv.most_similar([inferred_vector], topn=10)
sims
[(123, 0.9564133882522583),
(70, 0.9324836134910583),
(564, 0.9246233701705933),
(46, 0.9208188652992249),
(567, 0.9157204627990723),
(81, 0.9155042767524719),
(3, 0.9149802923202515),
(563, 0.9073331952095032),
(148, 0.9021652340888977),
(852, 0.9019087553024292)]
Consultando a partir do Google Acadêmico (GA)#
from scholarly import scholarly
import pandas as pd
No GA podemos facilmente buscar por autor. Por exemplo:
search_query = scholarly.search_author('Noam Chomsky')
author = next(search_query)
author
{'container_type': 'Author',
'filled': [],
'source': <AuthorSource.SEARCH_AUTHOR_SNIPPETS: 'SEARCH_AUTHOR_SNIPPETS'>,
'scholar_id': 'rbgNVw0AAAAJ',
'url_picture': 'https://scholar.google.com/citations?view_op=medium_photo&user=rbgNVw0AAAAJ',
'name': 'Noam Chomsky',
'affiliation': 'Professor of Linguistics (Emeritus), MIT',
'email_domain': '',
'interests': ['Linguistic Theory',
'Syntax',
'Semantics',
'Philosophy of Language'],
'citedby': 469642}
E então podemos listar suas publicações:
pubs = []
author = scholarly.fill(author) # Pegando mais detalhes
for pub in author['publications']:
# pub = scholarly.fill(pub)
pubs.append(pub)
dfpubs = pd.DataFrame(pubs)
dfpubs
| container_type | source | bib | filled | author_pub_id | num_citations | citedby_url | cites_id | |
|---|---|---|---|---|---|---|---|---|
| 0 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'Aspects of the Theory of Syntax', '... | False | rbgNVw0AAAAJ:znQEAXyEQe8C | 46180 | https://scholar.google.com/scholar?oi=bibs&hl=... | [1562037976341778186, 12173110054263842167] |
| 1 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'The minimalist program', 'pub_year'... | False | rbgNVw0AAAAJ:u-x6o8ySG0sC | 31996 | https://scholar.google.com/scholar?oi=bibs&hl=... | [8207792600950810812, 13969899446789541878] |
| 2 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'Syntactic structures', 'pub_year': ... | False | rbgNVw0AAAAJ:9yKSN-GCB0IC | 29918 | https://scholar.google.com/scholar?oi=bibs&hl=... | [2377703172325862585, 2983526281934768995, 485... |
| 3 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'Lectures on government and binding:... | False | rbgNVw0AAAAJ:uDl620nrkPcC | 28480 | https://scholar.google.com/scholar?oi=bibs&hl=... | [14028218170953085261, 4036591927768088961, 56... |
| 4 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'The sound pattern of English.', 'pu... | False | rbgNVw0AAAAJ:2osOgNQ5qMEC | 18935 | https://scholar.google.com/scholar?oi=bibs&hl=... | [1979786673247734672, 8099357462174375894] |
| ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 2526 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'EDITOR: Murray Eden EDITORIAL BOARD... | False | rbgNVw0AAAAJ:4MWp96NkSFoC | 0 | NaN | NaN |
| 2527 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'EINGEGANGENE SCHRIFTEN', 'citation'... | False | rbgNVw0AAAAJ:BUYA1_V_uYcC | 0 | NaN | NaN |
| 2528 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'AJ Kfoury', 'citation': ''} | False | rbgNVw0AAAAJ:EkHepimYqZsC | 0 | NaN | NaN |
| 2529 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'EDITOR: Murray Eden', 'citation': ''} | False | rbgNVw0AAAAJ:yB1At4FlUx8C | 0 | NaN | NaN |
| 2530 | Publication | PublicationSource.AUTHOR_PUBLICATION_ENTRY | {'title': 'Cumulatiw Contents of Voluares II t... | False | rbgNVw0AAAAJ:9Nmd_mFXekcC | 0 | NaN | NaN |
2531 rows × 8 columns