Sholhui Park
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Sholhui Park has a total of 58 co-authors. Here are the top co-authors by number of shared publications:
- Jungwon Huh
Department of Laboratory Medicine, Ewha Womans University College of Medicine, Seoul, Korea
14 shared publications Hae-Sun Chung
Department of Laboratory Medicine, College of Medicine, Ewha Womans University, Seoul, Republic of Korea
6 shared publications- Hae-Sun Chung
Department of Laboratory Medicine, Ewha Womans University College of Medicine, Seoul, Korea
6 shared publications Chu-Myong Seong
Department of Internal Medicine, College of Medicine, Ewha Womans University, Seoul, Republic of Korea
5 shared publications- Miae Lee
Department of Laboratory Medicine, Ewha Womans University College of Medicine, Seoul, Korea
5 shared publications - Min-Kyung So
Department of Laboratory Medicine, Ewha Womans University College of Medicine, Seoul, Korea
5 shared publications
Latest Publications
- 6 May 2026
- Published by Wiley in Vox Sanguinis
- Vol. 121 (8) , 1192-1200
An automatic consult reply system for therapeutic plasma exchange using retrieval‐augmented generation
Background and Objectives: Large language models (LLMs) show promise for clinical decision support but remain vulnerable to factual errors. Retrieval‐augmented generation (RAG) mitigates this limitation by grounding outputs in authoritative domain knowledge. Therapeutic plasma exchange (TPE) requires consistent, guideline‐driven decisions based on the 2023 American Society for Apheresis (ASFA) recommendations. This study aimed to evaluate whether an RAG‐based framework could improve accuracy, reliability and standardization of decision support for TPE, compared to conventional LLMs. Materials and Methods: We built a hybrid RAG pipeline combining BAAI/bge‐base‐en‐v1.5 embeddings with Chroma and BM25, coupled with structured prompts that encode ASFA categories and grades, Health Insurance Review and Assessment (HIRA) service criteria, and plasma volume computation rules. Thirty de‐identified real‐world consultation cases were converted into standardized queries. Across six RAG and three non‐RAG generative pre‐trained transformer (GPT)‐series model configurations, each case was answered five times (1,350 outputs). Performance was assessed by item‐level accuracy for six elements (diagnosis, ASFA category, grade, insurance applicability, plasma volume, and replacement fluid) and reproducibility on 14 disease‐name prompts. Response time and output length were also analyzed. Results: RAG configurations consistently outperformed non‐RAG baselines across items, with the largest gains in plasma‐volume calculation and ASFA classification. Reproducibility was markedly higher with RAG across repeated runs. Among all configurations, RAG GPT‐4.1‐mini showed the most balanced and superior performance, delivering high accuracy with low latency. Conclusion: A guideline‐grounded RAG approach substantially enhances the accuracy, stability and standardization of TPE consultation compared with conventional LLMs. This RAG‐TPE framework demonstrates the feasibility of reliable, clinically oriented decision support in transfusion medicine, warranting further evaluation in prospective clinical workflows.
- 6 May 2026
- Published by Wiley in Vox Sanguinis
- Vol. 121 (8) , 1192-1200
An automatic consult reply system for therapeutic plasma exchange using retrieval‐augmented generation
Background and Objectives: Large language models (LLMs) show promise for clinical decision support but remain vulnerable to factual errors. Retrieval‐augmented generation (RAG) mitigates this limitation by grounding outputs in authoritative domain knowledge. Therapeutic plasma exchange (TPE) requires consistent, guideline‐driven decisions based on the 2023 American Society for Apheresis (ASFA) recommendations. This study aimed to evaluate whether an RAG‐based framework could improve accuracy, reliability and standardization of decision support for TPE, compared to conventional LLMs. Materials and Methods: We built a hybrid RAG pipeline combining BAAI/bge‐base‐en‐v1.5 embeddings with Chroma and BM25, coupled with structured prompts that encode ASFA categories and grades, Health Insurance Review and Assessment (HIRA) service criteria, and plasma volume computation rules. Thirty de‐identified real‐world consultation cases were converted into standardized queries. Across six RAG and three non‐RAG generative pre‐trained transformer (GPT)‐series model configurations, each case was answered five times (1,350 outputs). Performance was assessed by item‐level accuracy for six elements (diagnosis, ASFA category, grade, insurance applicability, plasma volume, and replacement fluid) and reproducibility on 14 disease‐name prompts. Response time and output length were also analyzed. Results: RAG configurations consistently outperformed non‐RAG baselines across items, with the largest gains in plasma‐volume calculation and ASFA classification. Reproducibility was markedly higher with RAG across repeated runs. Among all configurations, RAG GPT‐4.1‐mini showed the most balanced and superior performance, delivering high accuracy with low latency. Conclusion: A guideline‐grounded RAG approach substantially enhances the accuracy, stability and standardization of TPE consultation compared with conventional LLMs. This RAG‐TPE framework demonstrates the feasibility of reliable, clinically oriented decision support in transfusion medicine, warranting further evaluation in prospective clinical workflows.
- 22 August 2025
- Published by Springer Nature in Journal of Community Genetics
- Vol. 16 (6) , 787-797
What kind of information is requested by patients and families with genetic disorders?
Background Genetic counseling is essential for patients and families with genetic disorders, providing accurate information and supporting informed decisions. However, limited access to counseling services in some countries can lead to confusion and anxiety, prompting many to seek information in online communities. This study analyzes user-generated questions from an online community in South Korea to understand the specific information needs of patients and families with genetic disorders. Methods This study analyzed 289 questions posted by 122 members on the Naver cafe < All About Genetic Disorders>(https://cafe.naver.com/geneticdx) between November 27, 2022, and December 23, 2023. Quantitative analysis was performed to identify the types and frequencies of questions, while qualitative analysis examined detailed content. Results The most frequently requested information was about disease information (28.4%), followed by genetics knowledge (26.6%), genetic testing (26.3%), and other topics (18.7%). Qualitative analysis revealed that patients and families needed detailed information about long-term progression and symptom manifestation. Many expressed confusion and anxiety regarding the meaning of variants of uncertain significance (VUS) in genetic testing results. They sought real-life patient experiences, in-depth professional informations, and wanted to know how to efficiently find accurate information. Conclusion This study demonstrated the importance of providing patients and families with professional and easily understandable information, highlighting the necessity for a well-organized genetic counseling system. To support patients and their families, it is essential to develop patient-friendly online platforms and expand access to genetic counseling services.
- 22 August 2025
- Published by Springer Nature in Journal of Community Genetics
- Vol. 16 (6) , 787-797
What kind of information is requested by patients and families with genetic disorders?
Background Genetic counseling is essential for patients and families with genetic disorders, providing accurate information and supporting informed decisions. However, limited access to counseling services in some countries can lead to confusion and anxiety, prompting many to seek information in online communities. This study analyzes user-generated questions from an online community in South Korea to understand the specific information needs of patients and families with genetic disorders. Methods This study analyzed 289 questions posted by 122 members on the Naver cafe < All About Genetic Disorders>(https://cafe.naver.com/geneticdx) between November 27, 2022, and December 23, 2023. Quantitative analysis was performed to identify the types and frequencies of questions, while qualitative analysis examined detailed content. Results The most frequently requested information was about disease information (28.4%), followed by genetics knowledge (26.6%), genetic testing (26.3%), and other topics (18.7%). Qualitative analysis revealed that patients and families needed detailed information about long-term progression and symptom manifestation. Many expressed confusion and anxiety regarding the meaning of variants of uncertain significance (VUS) in genetic testing results. They sought real-life patient experiences, in-depth professional informations, and wanted to know how to efficiently find accurate information. Conclusion This study demonstrated the importance of providing patients and families with professional and easily understandable information, highlighting the necessity for a well-organized genetic counseling system. To support patients and their families, it is essential to develop patient-friendly online platforms and expand access to genetic counseling services.
- Open Access
- 28 April 2025
- Published by Annals of Laboratory Medicine in Annals of Laboratory Medicine
- Vol. 45 (5) , 520-529
Evaluation of Six Large Language Models for Clinical Decision Support: Application in Transfusion Decision-making for RhD Blood-type Patients
Jong Kwon Lee , M.D., Sooin Choi , M.D., Sholhui Park , M.D., Sang-Hyun Hwang , M.D., Ph.D., and Duck Cho , M.D., Ph.D.. Ann Lab Med -0001;0:. https://doi.org/10.3343/alm.2024.0588
- 28 April 2025
- Published by Annals of Laboratory Medicine in Annals of Laboratory Medicine
- Vol. 45 (5) , 520-529
Open AccessEvaluation of Six Large Language Models for Clinical Decision Support: Application in Transfusion Decision-making for RhD Blood-type Patients
Jong Kwon Lee , M.D., Sooin Choi , M.D., Sholhui Park , M.D., Sang-Hyun Hwang , M.D., Ph.D., and Duck Cho , M.D., Ph.D.. Ann Lab Med -0001;0:. https://doi.org/10.3343/alm.2024.0588
- Open Access
- 14 December 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (24) , 2821
Reinterpretation of Conflicting ClinVar BRCA1 Missense Variants Using VarSome and CanVIG-UK Gene-Specific Guidance
Background: The accurate interpretation of the BRCA1/2 variant is critical for diagnosing and treating hereditary breast and ovarian cancers. ClinVar is a widely used public database for genetic variants. Conflicting classifications of pathogenicity can occur when different submitters categorize the same genetic variant inconsistently as pathogenic (PV), likely pathogenic (LPV), likely benign (LBV), benign (BV), or a variant of uncertain significance (VUS). The conflicting ClinVar BRCA1/2 variant classifications hinder clinical decision making. We reinterpreted 450 BRCA1 missense variants with conflicting interpretations in ClinVar (accessed on 20 December 2022). Methods: VarSome and the BRCA1/BRCA2: CanVIG-UK gene-specific guidance (CanVIG-UK) classifications were compared, and the five original classifications were consolidated into three categories (PV/LPV, VUS, and BV/LBV). Consensus analysis was performed between re-extracted ClinVar data and VarSome and CanVIG-UK results. Results: The three-category classification of the variants resulted in an overall concordance rate of 58.9% for BRCA1 missense variant interpretation between CanVIG-UK and VarSome, with VarSome having rates of 11.3, 24.7, and 64.0% for PV/LPV, VUS, and BV/LBV classifications and CanVIG-UK having rates of 11.1, 51.6, and 37.3% for P/LPV, VUS, and BV/LBV classifications, respectively. No variants classified as PV/LPV in VarSome were classified as BV/LBV in CanVIG-UK and vice versa. By 1 May 2024, 3.8% (17/450) of these conflicting variants reached a consensus classification in ClinVar and were definitively classified (9 PV/LPV, 1 VUS, and 7 BV/LBV). Conclusions: VarSome and CanVIG-UK have different features that help improve the accuracy of pathogenicity classification, highlighting the potential complementary use of both tools to support clinical decision making.
- 14 December 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (24) , 2821
Open AccessReinterpretation of Conflicting ClinVar BRCA1 Missense Variants Using VarSome and CanVIG-UK Gene-Specific Guidance
Background: The accurate interpretation of the BRCA1/2 variant is critical for diagnosing and treating hereditary breast and ovarian cancers. ClinVar is a widely used public database for genetic variants. Conflicting classifications of pathogenicity can occur when different submitters categorize the same genetic variant inconsistently as pathogenic (PV), likely pathogenic (LPV), likely benign (LBV), benign (BV), or a variant of uncertain significance (VUS). The conflicting ClinVar BRCA1/2 variant classifications hinder clinical decision making. We reinterpreted 450 BRCA1 missense variants with conflicting interpretations in ClinVar (accessed on 20 December 2022). Methods: VarSome and the BRCA1/BRCA2: CanVIG-UK gene-specific guidance (CanVIG-UK) classifications were compared, and the five original classifications were consolidated into three categories (PV/LPV, VUS, and BV/LBV). Consensus analysis was performed between re-extracted ClinVar data and VarSome and CanVIG-UK results. Results: The three-category classification of the variants resulted in an overall concordance rate of 58.9% for BRCA1 missense variant interpretation between CanVIG-UK and VarSome, with VarSome having rates of 11.3, 24.7, and 64.0% for PV/LPV, VUS, and BV/LBV classifications and CanVIG-UK having rates of 11.1, 51.6, and 37.3% for P/LPV, VUS, and BV/LBV classifications, respectively. No variants classified as PV/LPV in VarSome were classified as BV/LBV in CanVIG-UK and vice versa. By 1 May 2024, 3.8% (17/450) of these conflicting variants reached a consensus classification in ClinVar and were definitively classified (9 PV/LPV, 1 VUS, and 7 BV/LBV). Conclusions: VarSome and CanVIG-UK have different features that help improve the accuracy of pathogenicity classification, highlighting the potential complementary use of both tools to support clinical decision making.
- Open Access
- 4 April 2024
- Published by Annals of Laboratory Medicine in Annals of Laboratory Medicine
- Vol. 44 (6) , 478-486
Integration of an MC-80 Digital Image Analyzer With an Automated BC-6800Plus Hematology Analyzer Enables Accurate Platelet Counting in Samples With EDTA-Induced Pseudothrombocytopenia
Min-Kyung So, M.D., Jungwon Huh, M.D., Seunghwan Kim, M.D., and Sholhui Park, M.D.. Ann Lab Med -0001;0:. https://doi.org/10.3343/alm.2023.0460
- 4 April 2024
- Published by Annals of Laboratory Medicine in Annals of Laboratory Medicine
- Vol. 44 (6) , 478-486
Open AccessIntegration of an MC-80 Digital Image Analyzer With an Automated BC-6800Plus Hematology Analyzer Enables Accurate Platelet Counting in Samples With EDTA-Induced Pseudothrombocytopenia
Min-Kyung So, M.D., Jungwon Huh, M.D., Seunghwan Kim, M.D., and Sholhui Park, M.D.. Ann Lab Med -0001;0:. https://doi.org/10.3343/alm.2023.0460
- Open Access
- 8 October 2021
- Published by MDPI AG in Diagnostics
- Vol. 11 (10) , 1854
Variant Prediction by Analyzing RdRp/S Gene Double or Low Amplification Pattern in Allplex SARS-CoV-2 Assay
The spread of delta variants (B.1.671.2) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a severe global threat. Multiplex real-time PCR is a common method for confirming SARS-CoV-2 infection, however, additional tests, such as whole genomic sequencing, are required to reveal the presence or type of viral mutation. Moreover, applying whole genomic sequencing to all SARS-CoV-2 positive samples is challenging due to time and cost constraints. Here, we report that the double or low amplification curve observed during RNA-dependent RNA polymerase (RdRp) gene/S gene amplification in the Allplex SARS-CoV-2 Assay is related to delta/alpha variants. We analyzed the RdRp/S gene amplification curve using 94 samples confirmed as SARS-CoV-2 infection by the Allplex SARS-CoV-2 Assay from January to August, 2021. These positive samples identified variant types using the Novaplex SARS-CoV-2 Variants I and IV Assays. Overall, 17 samples showing a double curve and 11 samples showing a low amplification pattern were associated with alpha-/delta-type strains with variants in the P681 region. The double or low curve shown in the RdRp gene amplification curve had 100% sensitivity and 100% specificity for diagnosing delta/alpha variants. During the SARS-CoV-2 virus diagnostic RT-PCR test using the Allplex SARS-CoV-2 Assay, we could consider the presence of delta/alpha variants in the samples with double or low amplification curve of the RdRp/S gene channel. This PCR amplification curve abnormality enables rapid and cost-effective variant type prediction during SARS-CoV-2 diagnostic testing in clinical laboratories.
- 8 October 2021
- Published by MDPI AG in Diagnostics
- Vol. 11 (10) , 1854
Open AccessVariant Prediction by Analyzing RdRp/S Gene Double or Low Amplification Pattern in Allplex SARS-CoV-2 Assay
The spread of delta variants (B.1.671.2) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a severe global threat. Multiplex real-time PCR is a common method for confirming SARS-CoV-2 infection, however, additional tests, such as whole genomic sequencing, are required to reveal the presence or type of viral mutation. Moreover, applying whole genomic sequencing to all SARS-CoV-2 positive samples is challenging due to time and cost constraints. Here, we report that the double or low amplification curve observed during RNA-dependent RNA polymerase (RdRp) gene/S gene amplification in the Allplex SARS-CoV-2 Assay is related to delta/alpha variants. We analyzed the RdRp/S gene amplification curve using 94 samples confirmed as SARS-CoV-2 infection by the Allplex SARS-CoV-2 Assay from January to August, 2021. These positive samples identified variant types using the Novaplex SARS-CoV-2 Variants I and IV Assays. Overall, 17 samples showing a double curve and 11 samples showing a low amplification pattern were associated with alpha-/delta-type strains with variants in the P681 region. The double or low curve shown in the RdRp gene amplification curve had 100% sensitivity and 100% specificity for diagnosing delta/alpha variants. During the SARS-CoV-2 virus diagnostic RT-PCR test using the Allplex SARS-CoV-2 Assay, we could consider the presence of delta/alpha variants in the samples with double or low amplification curve of the RdRp/S gene channel. This PCR amplification curve abnormality enables rapid and cost-effective variant type prediction during SARS-CoV-2 diagnostic testing in clinical laboratories.
- Open Access
- 8 September 2021
- Published by MDPI AG in Diagnostics
- Vol. 11 (9) , 1644
Coexistence of Primary Myelofibrosis and Paroxysmal Nocturnal Hemoglobinuria Clone with JAK2 V617F, U2AF1 and SETBP1 Mutations: A Case Report and Brief Review of Literature
Primary myelofibrosis (PMF) and paroxysmal nocturnal hemoglobinuria (PNH) are very rare diseases, respectively, and it is uncommon to have both diseases together. Mutational profiling using next-generation sequencing in PMF and PNH detected additional mutations associated with myeloid neoplasms, suggesting a step-wise clonal evolution. We present here a very rare case with PMF and PNH with JAK2 V617F, U2AF1 and SETBP1 mutations at the time of diagnosis. The combination of these two diseases and three genetic mutations is difficult to interpret at once. (i.e., the sequence of these two clonal diseases or the time points of acquiring these mutations). Our report suggests that when diagnosing or treating patients with PMF, it is necessary to keep in mind that PNH may be present at the same time or sometimes new. The genetic mutations simultaneously found in this patient require further research to elucidate the clinical significance and their genetic associations fully.
- 8 September 2021
- Published by MDPI AG in Diagnostics
- Vol. 11 (9) , 1644
Open AccessCoexistence of Primary Myelofibrosis and Paroxysmal Nocturnal Hemoglobinuria Clone with JAK2 V617F, U2AF1 and SETBP1 Mutations: A Case Report and Brief Review of Literature
Primary myelofibrosis (PMF) and paroxysmal nocturnal hemoglobinuria (PNH) are very rare diseases, respectively, and it is uncommon to have both diseases together. Mutational profiling using next-generation sequencing in PMF and PNH detected additional mutations associated with myeloid neoplasms, suggesting a step-wise clonal evolution. We present here a very rare case with PMF and PNH with JAK2 V617F, U2AF1 and SETBP1 mutations at the time of diagnosis. The combination of these two diseases and three genetic mutations is difficult to interpret at once. (i.e., the sequence of these two clonal diseases or the time points of acquiring these mutations). Our report suggests that when diagnosing or treating patients with PMF, it is necessary to keep in mind that PNH may be present at the same time or sometimes new. The genetic mutations simultaneously found in this patient require further research to elucidate the clinical significance and their genetic associations fully.
- Open Access
- 30 June 2021
- Published by Korean Association of Quality Assurance for Clinical Laboratory in Laboratory Medicine and Quality Assurance
- Vol. 43 (2) , 80-86
Improvement of the Accuracy of Hepatitis B Surface Antigen Testing Using a Confirmatory Test
Hee Jin Huh, Sholhui Park, Jungwon Huh, Miae Lee, and Soo-Kyung Kim. Lab Med Qual Assur 2021;43:80-6. https://doi.org/10.15263/jlmqa.2021.43.2.80
- 30 June 2021
- Published by Korean Association of Quality Assurance for Clinical Laboratory in Laboratory Medicine and Quality Assurance
- Vol. 43 (2) , 80-86
Open AccessImprovement of the Accuracy of Hepatitis B Surface Antigen Testing Using a Confirmatory Test
Hee Jin Huh, Sholhui Park, Jungwon Huh, Miae Lee, and Soo-Kyung Kim. Lab Med Qual Assur 2021;43:80-6. https://doi.org/10.15263/jlmqa.2021.43.2.80
- Open Access
- 1 January 2021
- Published by Taylor & Francis in Hematology
- Vol. 26 (1) , 904-913
Comprehensive DNA repair gene expression analysis and its prognostic significance in acute myeloid leukemia
Deficiency in DNA damage response (DDR) pathway and accumulation of DNA damage increases mutation rates resulting in genomic instability and eventually increases the risk of cancer. The aim of our ...
- 1 January 2021
- Published by Taylor & Francis in Hematology
- Vol. 26 (1) , 904-913
Open AccessComprehensive DNA repair gene expression analysis and its prognostic significance in acute myeloid leukemia
Deficiency in DNA damage response (DDR) pathway and accumulation of DNA damage increases mutation rates resulting in genomic instability and eventually increases the risk of cancer. The aim of our ...
- Open Access
- 31 July 2020
- Published by The Ewha Medical Journal, Ewha Womans University College of Medicine in Ewha Medical Journal
- Vol. 43 (3) , 43-48
Application of Six Sigma Metrics to Improve Quality Control for Point-of-care Glucose Testing
Sholhui Park, Wookeun Lee, Tae-Dong Jeong, Hae-Sun Chung, Ki-Sook Hong. Ewha Med J 2020;43:43-8. https://doi.org/10.12771/emj.2020.43.3.43
- 31 July 2020
- Published by The Ewha Medical Journal, Ewha Womans University College of Medicine in Ewha Medical Journal
- Vol. 43 (3) , 43-48
Open AccessApplication of Six Sigma Metrics to Improve Quality Control for Point-of-care Glucose Testing
Sholhui Park, Wookeun Lee, Tae-Dong Jeong, Hae-Sun Chung, Ki-Sook Hong. Ewha Med J 2020;43:43-8. https://doi.org/10.12771/emj.2020.43.3.43