Sarah Feldman

Sarah Feldman

Auto-generated profile, based on publicly available information

Sarah Feldman

Unité d'Epidémiologie des Maladies Emergentes, Institut Pasteur, Paris, France

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 January 2023 in Nature Communications
Verified
Journal Article
Open access
PDF file
SLSophie Lefèvre
Sophie Lefèvre
ENElisabeth Njamkepo
Elisabeth Njamkepo
SFSarah Feldman
Sarah Feldman
CRCorinne Ruckly
Corinne Ruckly
ICIsabelle Carle
Isabelle Carle
FWFrançois-Xavier Weill
François-Xavier Weill
Shigella sonnei, the main cause of bacillary dysentery in high-income countries, has become increasingly resistant to antibiotics. We monitored the antimicrobial susceptibility of 7121 S. sonnei isolates collected in France between 2005 and 2021. We detected a dramatic increase in the proportion of isolates simultaneously resistant to ciprofloxacin (CIP), third-generation cephalosporins (3GCs) and azithromycin (AZM) from 2015. Our genomic analysis of 164 such extensively drug-resistant (XDR) isolates identified 13 different clusters within CIP-resistant sublineage 3.6.1, which was selected in South Asia ∼15 years ago. AZM resistance was subsequently acquired, principally through IncFII (pKSR100-like) plasmids. The last step in the development of the XDR phenotype involved various extended-spectrum beta-lactamase genes (blaCTX-M-3, blaCTX-M-15, blaCTX-M-27, blaCTX-M-55, and blaCTX-M-134) carried by different plasmids (IncFII, IncI1, IncB/O/K/Z) or even integrated into the chromosome, and encoding resistance to 3GCs. This rapid emergence of XDR S. sonnei, including an international epidemic strain, is alarming, and good laboratory-based surveillance of shigellosis will be crucial for informed decision-making and appropriate public health action.

1 January 2021 in Clinical Gastroenterology and Hepatology
Verified
Journal Article
Closed access
KYKyoko Yoshida
Kyoko Yoshida
ADAlice Desbiolles
Alice Desbiolles
SFSarah F. Feldman
Sarah F. Feldman
SASang Hoon Ahn
Sang Hoon Ahn
EAEnagnon K. Alidjinou
Enagnon K. Alidjinou
yusuke shimakawayusuke shimakawa
yusuke shimakawa

16 March 2021 in JMIR Medical Informatics
Verified
Journal Article
Open access
JJJordan Jouffroy
Jordan Jouffroy
SFSarah F Feldman
Sarah F Feldman
ILIvan Lerner
Ivan Lerner
BRBastien Rance
Bastien Rance
ABAnita Burgun
Anita Burgun
ANAntoine Neuraz
Antoine Neuraz
Background: Information related to patient medication is crucial for health care; however, up to 80% of the information resides solely in unstructured text. Manual extraction is difficult and time-consuming, and there is not a lot of research on natural language processing extracting medical information from unstructured text from French corpora. Objective: We aimed to develop a system to extract medication-related information from clinical text written in French. Methods: We developed a hybrid system combining an expert rule–based system, contextual word embedding (embedding for language model) trained on clinical notes, and a deep recurrent neural network (bidirectional long short term memory–conditional random field). The task consisted of extracting drug mentions and their related information (eg, dosage, frequency, duration, route, condition). We manually annotated 320 clinical notes from a French clinical data warehouse to train and evaluate the model. We compared the performance of our approach to those of standard approaches: rule-based or machine learning only and classic word embeddings. We evaluated the models using token-level recall, precision, and F-measure. Results: The overall F-measure was 89.9% (precision 90.8; recall: 89.2) when combining expert rules and contextualized embeddings, compared to 88.1% (precision 89.5; recall 87.2) without expert rules or contextualized embeddings. The F-measures for each category were 95.3% for medication name, 64.4% for drug class mentions, 95.3% for dosage, 92.2% for frequency, 78.8% for duration, and 62.2% for condition of the intake. Conclusions: Associating expert rules, deep contextualized embedding, and deep neural networks improved medication information extraction. Our results revealed a synergy when associating expert knowledge and latent knowledge.

3 February 2019 in Journal of Viral Hepatitis
Verified
Journal Article
Open access
PDF file
ZMZameer Mohamed
Zameer Mohamed
JKJin U. Kim
Jin U. Kim
AMAlex Magesa
Alex Magesa
MKMabula Kasubi
Mabula Kasubi
SFSarah F. Feldman
Sarah F. Feldman
MLMaud Lemoine
Maud Lemoine
Summary: Blood transfusion is one of the most commonly relied upon therapies in sub‐Saharan Africa. Existing safeguards recommended include systematic screening for transfusion‐transmitted infections and restricted voluntary nonremunerated blood donor selection. We report the transfusion‐transmitted infection screening and notification practice at a large urban blood transfusion centre in Dar‐es‐Salaam, Tanzania. Between October 2016 and March 2017 anonymized records of all donors registered at the blood transfusion unit were accessed to retrospectively note demographic information, donor status, first‐time status, transfusion‐transmitted infection result and notification. 6402 consecutive donors were screened for transfusion‐transmitted infections; the majority were family/replacement blood donors (88.0%) and male (83.8%). Overall transfusion‐transmitted infections prevalence was 8.4% (95% CI 7.8‐9.1), with hepatitis B being the most prevalent infection (4.1% (95% CI 3.6‐4.6)). Transfusion‐transmitted infections were more common in family/replacement blood donors (9.0% (95% CI 8.3‐9.8)) as compared to voluntary nonremunerated blood donor (4.1% (95% CI 2.8‐5.7)). A minority of infected‐donors were notified of a positive result (8.5% (95% CI 6.3‐11.2)). Although transfusion‐transmitted infections are more prevalent among family/replacement blood donors, overall risk of transfusion‐transmitted infections across all groups is considerable. In addition, existing efforts to notify donors of a positive transfusion‐transmitted infection are poor. Future policies must focus on improving linkage to care for newly diagnosed patients with transfusion‐transmitted infections.

Verified
Journal Article
Open access
PDF file
RMRachid MAHMOUDI
Rachid MAHMOUDI
SFSarah Feldman
Sarah Feldman
AKAymric Kisserli
Aymric Kisserli
VDValérie Duret
Valérie Duret
TTThierry Tabary
Thierry Tabary
JCJacques Henri Max Cohen
Jacques Henri Max Cohen
The complement receptor 1 (CR1) gene was shown to be involved in Alzheimer’s disease (AD). We previously showed that AD is associated with low density of the long CR1 isoform, CR1*2 (S). Here, we correlated phenotype data (CR1 density per erythrocyte (CR1/E), blood soluble CR1 (sCR1)) with genetic data (density/length polymorphisms) in AD patients and healthy controls. CR1/E was enumerated using flow cytometry, while sCR1 was quantified by ELISA. CR1 polymorphisms were assessed using restriction fragment length polymorphism (RFLP), pyrosequencing, and high-resolution melting PCR. In AD patients carrying the H allele (HindIII polymorphism) or the Q allele (Q981H polymorphism), CR1/E was significantly lower when compared with controls carrying the same alleles (p < 0.01), contrary to sCR1, which was significantly higher (p < 0.001). Using multivariate analysis, a reduction of 6.68 units in density was associated with an increase of 1% in methylation of CR1 (estimate −6.68; 95% confidence intervals (CIs) −12.37, −0.99; p = 0.02). Our data show that, in addition to inherited genetic factors, low density of CR1/E is also acquired. The involvement of CR1 in the pathogenesis of AD might be linked to insufficient clearance of amyloid deposits. These findings may open perspectives for new therapeutic strategies in AD.

1 July 2015 in Critical Care Medicine
Verified
Journal Article
Closed access
CJCheng Jiang
Cheng Jiang
ACAude Carillion
Aude Carillion
NNNa Na
Na Na
AJAudrey De Jong
Audrey De Jong
SFSarah Feldman
Sarah Feldman
JAJulien Amour
Julien Amour
Although metabolic syndrome is associated with increased sympathetic activity that chronically stimulates β-adrenoceptors, the β-adrenoceptor signaling pathway has been poorly studied in this situation. We studied the β-adrenoceptor signaling pathway in Zucker lean, obese, and obese diabetic rats. Experimental, prospective study. University medical research laboratory. Adult male Zucker lean (control), obese, and obese diabetic rats. The effects of β-adrenoceptor stimulation were investigated in vitro in isolated left ventricular papillary muscles in control, obese, and obese diabetic rats. β1-, β2-, and β3-adrenoceptors and multidrug resistance-associated protein 4 were quantified by Western Blotting. Triglyceride, cholesterol, leptin, adiponectin, and C-peptide plasma concentrations were measured. Data are mean ± SD. Hyperlipidemia, high leptin, and C-peptide concentrations were observed in obese and obese diabetic strains, whereas hyperglycemia occurred only in the diabetic strain. The positive inotropic effect of isoproterenol was slightly reduced in obese rats (183% ± 11% of baseline; p = 0.003; n = 7) and markedly reduced in obese diabetic rats (137% ± 18% of baseline; p < 0.001; n = 10) when compared with control rats (210% ± 17% of baseline; n = 9). β1-adrenoceptors were down-regulated in obese (-41%; p = 0.02) and diabetic (-54%; p = 0.003) when compared with control rats, whereas β3-adrenoceptors and multidrug resistance-associated protein expression remained unchanged. Direct stimulation of adenylate cyclase with forskolin or administration of 3',5'-cyclic adenosine monophosphate suggests that subtle impairments also occurred beside the down-regulation of β1-adrenoceptor. The positive inotropic effect of β-adrenoceptor stimulation is slightly decreased in Zucker obese rats and was more markedly decreased in Zucker diabetic rats. These decreases are mainly related to β1-adrenoceptor down-regulation.