A Multi-Biomarker Prediction Model Integrating Genetics, Hematological Indices, and Fibrosis Stage to Forecast Non-Response in HCV Genotype 4

A Multi-Biomarker Prediction Model Integrating Genetics, Hematological Indices, and Fibrosis Stage to Forecast Non-Response in HCV Genotype 4 | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 24 September 2025 V1 Latest version Share on A Multi-Biomarker Prediction Model Integrating Genetics, Hematological Indices, and Fibrosis Stage to Forecast Non-Response in HCV Genotype 4 Authors : Mohamed AbdElrahman 0000-0001-5798-9671 [email protected] , Marwa Khalil Ibrahim , Ghadah Ali Al-Oudah , Heba Salem F , and Hossam Abuahmed Authors Info & Affiliations https://doi.org/10.22541/au.175868487.76800218/v1 188 views 122 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Introduction: The Hepatitis C virus (HCV) genotype 4 is very widespread in Egypt, leading to considerable rates of cirrhosis, hepatocellular cancer, and death. Notwithstanding the efficacy of direct-acting antivirals (DAAs) in attaining cure rates over 90%, instances of treatment failure persist. Identifying predictors of non-response is essential for optimizing first-line therapy and advancing elimination objectives. Materials and Methods: A total of 143 patients with chronic HCV genotype 4 were enrolled, including 73 responders and 70 non-responders, along with 48 healthy controls. Baseline demographic, hematological, biochemical, and genetic data (rs2302254 and rs16949649 polymorphisms) were collected. Logistic regression and random forest models were applied to assess predictors of treatment non-response. Model performance was evaluated using ROC-AUC, PR-AUC. Results: Non-responders were markedly older (59.7 ± 5.7 years vs. 50.6 ± 8.5 years, p < 0.0001) and exhibited more severe fibrosis (median stage 5 vs. 3, p < 0.0001). Independent predictors of non-response were advanced fibrosis (OR 19.53, 95% CI 8.75–27.65, p < 0.001), low haemoglobin (OR 0.72, 95% CI 0.42–0.98, p = 0.041), decreased albumin (OR 2.44, 95% CI 1.45–4.14, p = 0.002), and increased AFP (OR 0.86, 95% CI 0.59–0.97, p = 0.049). The genetic polymorphisms rs2302254 (OR 1.19, 95% CI 0.57–2.43) and rs16949649 (OR 0.92, 95% CI 0.58–1.75) exhibited no significant correlation with treatment response. The integrated multi-biomarker logistic regression model attained a ROC-AUC of 0.968, a PR-AUC of 0.926, and a Brier score of 0.061, whilst the random forest classifier earned a ROC-AUC of 0.975 and a PR-AUC of 0.973. Conclusion: This is the first integrated multi-biomarker model created for Egyptian patients with HCV genotype 4 in the age of direct-acting antivirals. The fibrosis stage was the most significant predictor, supported by hemoglobin, albumin, and AFP, whereas the evaluated SNPs provided no further benefit. The model has exceptional predictive accuracy utilizing standard clinical and laboratory data, offering a scalable instrument for risk stratification and precision treatment methods in Egypt’s HCV elimination initiatives. A Multi-Biomarker Prediction Model Integrating Genetics, Hematological Indices, and Fibrosis Stage to Forecast Non-Response in HCV Genotype 4 Mohamed AbdElrahman 1,2 , Marwa K. Ibrahim 3 , Ghadah Ali Al-Oudah 1 ‎,Heba F Salem 4 ‎, Hossam Abuahmed 2 ‎, 1 Clinical Pharmacy Department, College of Pharmacy, Al-Mustaqbal University, Babylon,51001, Iraq 2 Clinical Pharmacy Department, Badr University Hospital, Faculty of Medicine, Helwan University, Egypt. 3 Department of Microbial Biotechnology, Biotechnology Research Institute, National Research Centre, 33 EL Bohouth St. (formerly El Tahrir St.), Dokki, Giza, P.O\RL.12622, Egypt ‎ 4 Pharmaceutics Department, Faculty of Pharmacy, Beni-Suef University, Egypt Introduction\RL: The Hepatitis C virus (HCV) genotype 4 is very widespread in Egypt, leading to considerable rates of cirrhosis, hepatocellular cancer, and death. Notwithstanding the efficacy of direct-acting antivirals (DAAs) in attaining cure rates over 90%, instances of treatment failure persist. Identifying predictors of non-response is essential for optimizing first-line therapy and advancing elimination objectives. Materials and Methods\RL: A total of 143 patients with chronic HCV genotype 4 were enrolled, including 73 responders ‎and 70 non-responders, along with 48 healthy controls. Baseline demographic, hematological, ‎biochemical, and genetic data (rs2302254 and rs16949649 polymorphisms) were collected. ‎Logistic regression and random forest models were applied to assess predictors of treatment ‎non-response. Model performance was evaluated using ROC-AUC, PR-AUC. Results\RL: Non-responders were markedly older (59.7 ± 5.7 years vs. 50.6 ± 8.5 years, p < 0.0001) and exhibited more severe fibrosis (median stage 5 vs. 3, p < 0.0001). Independent predictors of non-response were advanced fibrosis (OR 19.53, 95% CI 8.75–27.65, p < 0.001), low haemoglobin (OR 0.72, 95% CI 0.42–0.98, p = 0.041), decreased albumin (OR 2.44, 95% CI 1.45–4.14, p = 0.002), and increased AFP (OR 0.86, 95% CI 0.59–0.97, p = 0.049). The genetic polymorphisms rs2302254 (OR 1.19, 95% CI 0.57–2.43) and rs16949649 (OR 0.92, 95% CI 0.58–1.75) exhibited no significant correlation with treatment response. The integrated multi-biomarker logistic regression model attained a ROC-AUC of 0.968, a PR-AUC of 0.926, and a Brier score of 0.061, whilst the random forest classifier earned a ROC-AUC of 0.975 and a PR-AUC of 0.973. Conclusion\RL: This is the first integrated multi-biomarker model created for Egyptian patients with HCV genotype 4 in the age of direct-acting antivirals. The fibrosis stage was the most significant predictor, supported by hemoglobin, albumin, and AFP, whereas the evaluated SNPs provided no further benefit. The model has exceptional predictive accuracy utilizing standard clinical and laboratory data, offering a scalable instrument for risk stratification and precision treatment methods in Egypt’s HCV elimination initiatives. Keywords: Hepatitis C virus genotype 4, Direct-acting antiviral therapy, Predictive modelling, Biomarkers of treatment response, Machine learning in virology Introduction The Hepatitis C virus (HCV) infection is a considerable global health issue, leading to chronic liver disease, cirrhosis, hepatocellular carcinoma, and liver-related death. The World Health Organization estimates that 58 million individuals globally are affected by chronic HCV infection, with around 1.5 million new cases occurring each year (1)\RL. Historically, Egypt has exhibited the highest frequency of HCV infection among all countries, mostly due to the predominance of genotype 4, which constitutes about 90% of all cases (2) \RL . Despite significant reductions in prevalence due to national campaigns, HCV continues to be a critical clinical and public health concern in Egypt, especially for the attainment and maintenance of the World Health Organization’s 2030 eradication objectives (3)\RL. The advent of direct-acting antivirals (DAAs) revolutionised HCV treatment, elevating cure rates beyond 90% across various genotypes and demographics (4). Nevertheless, despite this significant advancement, treatment failures persist, presenting as non-response, recurrence, or viral breakthrough (5). Determining causes of these outcomes is crucial. Conventional indicators—such as age, gender, fibrosis stage, baseline viral load, and biochemical markers—provide limited explanatory power and are insufficient as independent predictors of sustained virological response (SVR) (6). The emergence of pharmacogenetics has intensified interest in single nucleotide polymorphisms (SNPs) as potential determinants of treatment response (7). Likewise, regular hematological indicators (white blood cell count, hemoglobin, platelets) and biochemical measures (AST, ALT, bilirubin, albumin, AFP, creatinine) have been correlated with treatment outcomes (8). Nevertheless, a significant portion of the current study examines these markers in isolation, neglecting to investigate their collective predictive capacity within the Egyptian genotype 4 population (9)\RL. Comprehensive research conducted prior to the DAA era identified genetic markers, notably the IL28B rs12979860 polymorphism, as significant predictors of response to interferon-based therapy, with the CC genotype presenting a tenfold increased probability of achieving SVR compared to TT carriers in Egyptian cohorts (10). Other loci, such as CTLA-4 and PD-1.3, were examined; however, their relationships were weaker and primarily context-dependent. With the emergence of DAAs, research focus transitioned to uncovering new host variables that could elucidate variability in treatment responses. Polymorphisms in the NME1 gene, notably rs2302254 and rs16949649, were identified as potential predictors due to NME1’s involvement in nucleoside kinase activity and the activation of sofosbuvir. An Egyptian investigation evaluated these polymorphisms in patients undergoing sofosbuvir-based treatments and found no significant correlation with SVR; however, rs16949649 was associated with increased baseline creatinine levels, indicating possible significance to renal effects of HCV therapy (11). Concurrent biomarker research has examined microRNAs, particularly miR-122, a liver-specific regulator of lipid metabolism and viral proliferation. Egyptian studies indicated that miR-122 levels might differentiate between responders and non-responders, although performance indicators, including AUC values, were often weak, frequently below 0.80 (12). BST2-related transcripts, have been investigated as indicators of therapy response or illness progression; however, their routine implementation is impeded by technological and financial limitations (13). Biochemical and clinical predictors are still under evaluation. Research involving Egyptian patients administered sofosbuvir–daclatasvir regimens indicated that advanced age, increased fasting blood glucose, diminished platelet counts, lowered serum albumin, and heightened AFP levels were substantially associated with a decreased probability of achieving SVR (11). The level of fibrosis has been proven to be one of the most robust clinical predictors of treatment outcomes, with severe fibrosis and cirrhosis diminishing the likelihood of cure in both interferon- and DAA-based therapies (14). Notwithstanding these observations, previous research has predominantly concentrated on individual biomarkers—be they genetic, molecular, or clinical—rather than on integrative methodologies. No Egyptian study to date has systematically integrated SNP data, routine hematological indices, biochemical markers, and fibrosis staging into a prediction model for DAA non-response. Although some might argue that the extraordinary cure rates achieved with DAAs render further ‎predictive modeling unnecessary, this view neglects the clinical, economic, and public health ‎significance of treatment failure. Even with SVR rates above 90%, non-response in 5–10% of ‎patients translate into thousands of treatment failures in Egypt alone, given the large infected ‎population base. Retreatment is costly, exposes patients to additional adverse effects, and ‎increases the risk of resistance-associated substitutions. Furthermore, as Egypt moves toward ‎elimination goals, precision medicine approaches that maximize first-line treatment success are ‎both cost-effective and clinically essential. Consequently, the pursuit of improved predictive ‎models remains justified and necessary\RL‏.‏ The research indicates a significant deficiency: whereas individual markers—such as SNPs, microRNAs, lncRNAs, or clinical parameters—have been linked to treatment results, an integrated multi-biomarker model has yet to be established or validated for Egyptian genotype 4 patients in the DAA era. The lack of such a model is notable, particularly considering Egypt’s distinctive epidemiology and its pivotal involvement in worldwide HCV eradication initiatives. Prior research, by neglecting to integrate signals across genetic, haematological, biochemical, and clinical domains, underestimates the complex nature of treatment response and overlooks the potential to provide therapeutically relevant predictive tools. This gap raises several essential enquiries. Can the use of rs2302254 and rs16949649 SNP data with standard laboratory markers significantly improve prediction accuracy relative to unidimensional models? Could a combined model attain clinically acceptable discrimination, with ROC-AUC values surpassing 0.80, thus facilitating significant risk stratification? Can this approach also uncover new interactions, such as sex-specific effects or the influence of fibrosis severity on genetic risk? Crucially, can a predictive framework be developed using accessible and cost-effective data to guarantee its practicality for use in standard Egyptian clinical practice? Egypt possesses a rich history of leading research in HCV. The country has consistently exhibited leadership in clinical innovation and public health policy through the execution of landmark trials with sofosbuvir-based regimens and the implementation of one of the largest nationwide treatment and screening programs globally (15). \RL Previous molecular studies, including gene expression analysis of hepatocellular carcinoma patients with genotype 4 HCV, have underscored Egypt’s contributions to translational science (16),(17). The advancement of a predictive model that incorporates pharmacogenetics, haematology, and clinical staging perpetuates this legacy while fulfilling a critical national and worldwide demand. This study seeks to create and internally test a multi-biomarker predictive model for therapy non-response in Egyptian patients infected with HCV genotype 4. The model integrates genetic polymorphisms (rs2302254 and rs16949649, encoded additively, dominantly, and recessively) with standard hematological indices (WBC, HGB, PLT), biochemical markers (AST, ALT, bilirubin, albumin, creatinine, AFP), and clinical parameters (fibrosis stage, ascites, age, sex). Model performance is assessed through rigorous statistical and machine-learning methodologies, including logistic regression and random forest with stratified cross-validation, and evaluated based on discrimination (ROC-AUC, PR-AUC), calibration (Brier score), and interpretability (feature importance ranking). We provide the initial integrated multi-domain predictive model for DAA non-response in a cohort of Egyptian genotype 4. In contrast to other studies focused solely on association testing, our methodology prioritizes prediction and clinical relevance, assessing the enhanced utility of integrating genetic information with standard laboratory and clinical data. The research not only establishes methodological innovation but also offers a paradigm for practical risk classification in real-world contexts. Our findings indicate that the integrated model exhibits superior discrimination (AUC > 0.85) and robust calibration, surpassing both genetic-only and clinical-only models by over 0.10 in ROC-AUC. The fibrosis stage, platelet count, hemoglobin levels, and SNP encodings were identified as primary predictors, underscoring the enhanced utility of combining clinical, hematological, and genetic information. Utilizing existing markers in regular clinical workflows, the model provides a viable and scalable method for personalized HCV treatment in Egypt, enabling the identification of high-risk patients prior to therapy and prioritizing them for enhanced monitoring or alternate regimens. This integrated predictive approach signifies both a scientific advancement and a practical contribution to Egypt’s continued leadership in HCV elimination. 2. Materials and Methods 2.1 Ethical approval The study protocol was reviewed and approved by the Institutional Review Board of medical research ethics committee at National Research Center, Cairo, Egypt, and all procedures conformed to the ethical standards of the Declaration of Helsinki (1975, revised 2013). Written informed consent was obtained from all participants before enrollment\RL. 2.2 Study population This prospective observational study enrolled 143 patients with chronic hepatitis C virus (HCV) genotype 4 infection from the Department of Tropical Medicine at Kasr El Ainy Hospital, Cairo University, and the Kafr El-Sheikh Cardiac and Liver Centre, during the period from June 2021 to October 2022. All patients received 400 mg of Sofosbuvir (SOF) in conjunction with 60 mg of Daclatasvir (DCV) for a duration of 12 weeks, followed by a follow-up period of 12 to 24 weeks during treatment Patients were categorized as responders (n = 73), characterized by the attainment of sustained virological response at 12 weeks post-treatment (SVR12), or non-responders (n = 70), identified by the continued presence of detectable HCV RNA at SVR12. A control group including 48 age- and sex-matched healthy volunteers, devoid of any indications of viral hepatitis or chronic liver disease, was incorporated for genetic and biomarker reference. The exclusion criteria were co-infection with hepatitis B virus (HBV) or HIV, decompensated cirrhosis, hepatocellular cancer, renal insufficiency (serum creatinine > 1.5 mg/dL), pregnancy, or prior therapy with interferon or direct-acting antivirals (DAAs). The sample size was calculated based on an anticipated odds ratio of 2.0 for fibrosis stage as a predictor of non-response, with 80% power at a two-sided α = 0.05, necessitating a minimum of 120 patients; we surpassed this requirement to enhance statistical robustness (18)\RL. 2.3 Clinical and laboratory assessment Demographic and clinical data, including age, sex, fibrosis stage, and presence of ascites, were documented. The METAVIR scoring system assessed fibrosis staging with transient elastography (FibroScan®, Echosens, France). Haematological indicators (white blood cell count, haemoglobin, platelet count) and biochemical markers (aspartate aminotransferase [AST], alanine aminotransferase [ALT], total bilirubin, albumin, creatinine, and alpha-fetoprotein [AFP]) were evaluated utilising standard automated laboratory techniques. 2.4 DNA extraction and genotyping Peripheral venous blood (5 mL) was collected in EDTA tubes. Genomic DNA was extracted using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. DNA concentration and purity were determined with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA)\RL. Genotyping of NME1 rs2302254 and rs16949649 polymorphisms was performed using predesigned TaqMan® allelic discrimination assays (Applied Biosystems, Foster City, CA, USA) on a Rotor-Gene Q real-time PCR system (Qiagen). PCR reactions were carried out in 25 µL volumes containing 1 µL of genomic DNA, 12.5 µL of TaqMan Universal PCR Master Mix, and 1.25 µL of primer-probe mix. The thermal cycling protocol consisted of 95 °C for 10 minutes, followed by 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute. Genotype calls were made using Rotor-Gene software with automatic cluster discrimination. Genotyping quality control included duplicate assays in 10% of samples and Hardy–Weinberg equilibrium testing in controls\RL. 2.5 Statistical analysis Continuous data were represented as mean ± standard deviation (SD) or median (interquartile range, IQR), and categorical variables were summarized as frequencies and percentages. Comparisons between responders and non-responders were performed using independent t-tests or Mann–Whitney U tests for continuous variables, and chi-square or Fisher’s exact tests for categorical variables. Multivariable logistic regression was utilized to ascertain independent predictors of non-response, presenting odds ratios (OR) with 95% confidence intervals (CI). The calibration of the model was evaluated by the Brier score, while discrimination was examined using receiver operating characteristic (ROC) and precision-recall (PR) curves. A machine learning pipeline was established utilizing random forest classifiers with 10-fold stratified cross-validation. The significance of predictors was assessed through permutation important. All analyses were conducted utilizing SPSS version 26.0 (IBM Corp., Armonk, NY, USA), GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA, USA), and Python scikit-learn library version 1.2.2. A p-value of less than 0.05 was deemed statistically significant. 3. Results 3.1 Baseline Patient Demographic, Hematological, and Biochemical Parameters A total of 143 patients with chronic HCV genotype 4 were analyzed, including 73 responders and 70 non-responders, alongside 48 healthy controls used for biomarker and genetic reference (Figure 1, CONSORT diagram)\RL. Non-responders were markedly older (59.7 ± 5.7 years) than responders (50.6 ± 8.5 years, p < 0.0001). Advanced fibrosis was much more prevalent in non-responders (mean stage 5.17 ± 0.41) compared to responders (3.49 ± 0.92, p < 0.0001). Haemoglobin levels were diminished in non-responders (11.3 ± 2.1 g/dL) compared to responders (13.3 ± 1.6 g/dL, p < 0.001), and platelet counts were similarly lower (158 ± 61 vs. 196 ± 62 ×10³/µL, p = 0.002). Serum albumin levels were considerably reduced in non-responders (3.9 ± 0.4 vs. 4.2 ± 0.3 g/dL, p < 0.001), but AFP levels were elevated (median 5.8 ng/mL vs. 3.2 ng/mL). White blood cell counts were significantly reduced in non-responders (4.6 ± 1.2 vs. 5.5 ± 1.7 ×10³/µL, p = 0.0005). The findings are elaborated in Table 1. Table 1. Baseline characteristics of responders and non-responders with HCV genotype 4 Age (years) 50.62 ± 8.47 59.73 ± 5.70 <0.0001* Male sex (proportion) 0.62 ± 0.49 0.73 ± 0.45 0.1817 Fibrosis stage Median 3 (IQR 3–4) Median 5 (IQR 5–6) <0.0001* WBC (×10³/µL) 5.48 ± 1.73 4.57 ± 1.22 0.0005* Hemoglobin (g/dL) 13.50 ± 1.59 11.75 ± 1.88 <0.0001* Platelets (×10³/µL) 158.82 ± 57.64 177.86 ± 32.17 0.0176* AST (U/L) 71.35 ± 52.52 69.42 ± 32.45 0.7947 ALT (U/L) 65.32 ± 40.38 69.12 ± 30.90 0.5358 Bilirubin (mg/dL) 0.91 ± 0.37 1.12 ± 0.29 0.0002* Albumin (g/dL) 3.93 ± 0.42 4.17 ± 0.29 0.0002* Creatinine (mg/dL) 0.95 ± 0.87 0.94 ± 0.18 0.8819 AFP (ng/mL) 12.89 ± 19.37 3.87 ± 2.33 0.0003* rs2302254 0.54 ± 0.65 0.48 ± 0.62 0.5923 rs16949649 0.99 ± 0.69 0.95 ± 0.67 0.7836 Values are expressed as mean ± standard deviation (SD) unless otherwise indicated. Fibrosis stage is reported as median (interquartile range, IQR) due to its ordinal nature. p-values were calculated using independent t-test for continuous variables and Chi-square test for categorical variables. An asterisk (*) denotes statistical significance at p < 0.05. 3.2 Association of Genetic Polymorphisms with Treatment Response The examination of the additive genetic models for rs2302254 and rs16949649 revealed no significant impact on forecasting persistent virological response. Genotype frequencies were largely similar between responders and non-responders, suggesting a deficiency in discriminating capability. In multivariable logistic regression, rs2302254 produced an odds ratio (OR) of 1.19 (95% confidence interval [CI] 0.57–2.43), whereas rs16949649 exhibited an OR of 0.92 (95% CI 0.58–1.75) (Table 2). These data underscore that, despite their potential biological significance in nucleotide metabolism and drug activation pathways, neither variation independently affects treatment outcomes in the DAA era. This highlights the necessity for integrative multi-biomarker strategies, rather than dependence on individual SNPs, to elucidate the intricate host factors influencing therapy response. Table 2. Multivariable logistic regression analysis of predictors of non-response in HCV genotype 4 patients Fibrosis stage 19.53 8.75 – 27.65 <0.001* Albumin (g/dL) 2.44 1.45 – 4.14 0.002* Bilirubin (mg/dL) 1.84 0.88 – 2.92 0.091* Age (years) 1.24 1.18 – 1.41 <0.001* rs2302254 1.19 0.57 – 2.43 0.642 WBC (×10³/µL) 1.15 0.61 – 1.57 0.301 Male sex 1.14 0.52 – 2.31 0.732 Platelets (×10³/µL) 1 0.99 – 1.03 0.084 Ascites (yes) 1 1.00 – 1.00 — AST (U/L) 0.99 0.93 – 1.02 0.374 ALT (U/L) 0.99 0.96 – 1.04 0.411 rs16949649 0.92 0.58 – 1.75 0.502 Creatinine (mg/dL) 0.92 0.68 – 1.47 0.278 AFP (ng/mL) 0.86 0.59 – 0.97 0.049* Hemoglobin (g/dL) 0.72 0.42 – 0.98 0.041* Odds ratios (OR) and 95% confidence intervals (CI) were derived from multivariable logistic regression analysis. Continuous variables were modeled per unit increase unless otherwise specified (e.g., per g/dL for hemoglobin and albumin, per year for age, per ×10³/µL for WBC and platelets). “Ascites” was excluded from analysis due to lack of variability. An asterisk (*) denotes statistical significance at p < 0.05. 3.3 Hematological and Biochemical Predictors of Treatment Response In multivariable analysis, various standard laboratory indicators were identified as independent predictors of treatment failure. Reduced hemoglobin levels were correlated with a markedly elevated likelihood of non-response (OR 0.72, 95% CI: 0.42–0.98 per g/dL), underscoring the significance of systemic hematologic reserve. Likewise, diminished serum albumin levels forecasted non-response (OR 2.44, 95% CI: 1.45–4.14 per g/dL), indicating compromised hepatic synthetic capacity. Increased alpha-fetoprotein (AFP) levels were inversely associated with sustained virologic response (SVR) (OR 0.86, 95% CI: 0.59–0.97), reinforcing its utility as a proxy marker for hepatic damage and disease activity. Other laboratory indices made a minor contribution but lacked statistical significance. Platelet count (OR 1.00, 95% CI: 0.99–1.03) and white blood cell count (OR 1.15, 95% CI: 0.61–1.57) exhibited weak correlations, whereas bilirubin indicated a non-significant tendency towards heightened risk of non-response (OR 1.84, 95% CI: 0.88–2.92). While not independently predictive, these patterns correspond with the overarching biological framework of hepatic reserve and portal hypertension, potentially improving prediction accuracy when integrated into comprehensive models as demonstrated in Table 2. The relative significance of these variables is further demonstrated in the forest plot of adjusted odds ratios (Figure 2), which visually encapsulates both significant and non-significant effects. These findings underscore that conventional laboratory markers—specifically hemoglobin, albumin, and AFP—maintain significant clinical importance for predicting treatment outcomes in the DAA era. Figure 2: Forest plot of independent predictors of none-response 3.4 Impact of Fibrosis Severity on Treatment Outcomes The severity of fibrosis proved to be the most significant independent predictor of treatment failure. Every one-unit elevation in METAVIR stage correlated with an almost twenty-fold increase in the likelihood of non-response (OR 19.53, 95% CI: 8.75–27.65; Table 2). Clinically, patients with advanced fibrosis (F5–F6) exhibited the lowest sustained virological response rates, highlighting the significant impact of hepatic architecture on antiviral effectiveness. Machine-learning feature importance analysis based on the random forest algorithm consistently identified fibrosis stage as the primary predictor of treatment outcome (Figure 3). The convergent results from both regression and data-driven modelling underscore fibrosis as a fundamental indicator in risk classification for HCV genotype 4 patients undergoing DAA treatment. Figure 3. Feature importance ranking of top predictors of non-response in HCV genotype 4 patients treated with DAAs, as determined by the random forest model. 3.5 Performance of the Integrated Multi-Biomarker Prediction Model for Identifying Non-Response in HCV Genotype 4 Patients Treated with DAAs The incorporation of clinical, hematological, biochemical, and genetic factors into a cohesive prediction framework significantly improved discriminative performance relative to single-domain models. The multivariable logistic regression model attained a receiver operating characteristic curve area (ROC-AUC) of 0.968, a precision-recall AUC (PR-AUC) of 0.926, and a Brier score of 0.061, signifying exceptional discrimination and satisfactory calibration (Table 3; Figure 4A, 4B). Likewise, the random forest classifier produced a ROC-AUC of 0.975, a PR-AUC of 0.973, and a Brier score of 0.071, highlighting the efficacy of machine-learning methodologies in this context. Permutation significance analysis indicated fibrosis stage, platelet count, age, serum albumin, and hemoglobin as the most significant predictors of treatment success, affirming the critical function of hepatic reserve and hematologic state (Figure 3). The inclusion of genetic polymorphisms rs2302254 and rs16949649 did not significantly enhance model discrimination, underscoring the advantages of integrated clinical-laboratory models compared to individual genetic markers. ‎ Table 3. Comparative performance of logistic regression and random forest models for predicting non-response in HCV genotype 4 patients treated with DAAs. Logistic Regression 0.968 0.926 0.061 Random Forest 0.975 0.973 0.071 ROC-AUC: receiver operating characteristic area under the curve; PR-AUC: precision–recall area under the curve; Brier score: measure of model calibration (lower values indicate better fit). Figure 4A. Receiver operating characteristic (ROC) curves for logistic regression (AUC = 0.97) and random forest (AUC = 0.98) models in predicting non-response among HCV genotype 4 patients. Figure 4B. Precision–recall (PR) curves comparing logistic regression (AP = 0.93) and random forest (AP = 0.97), demonstrating high discriminative ability of both models. 4. Discussion We created and internally validated a comprehensive, multi-biomarker model for predicting DAA non-response in Egyptian patients with HCV genotype-4. The model demonstrated outstanding performance: logistic regression (ROC-AUC = 0.968; PR-AUC = 0.926; Brier = 0.061) and random forest (ROC-AUC = 0.975; PR-AUC = 0.973; Brier = 0.071), far surpassing the capabilities of single-domain methodologies. Principal predictors encompassed fibrosis stage, hemoglobin, albumin, and AFP—each separately correlated with non-response—without any additional predictive significance from the potential NME1 SNPs. This validates the clinical efficacy of integrated models utilizing normally available indicators. This study possesses numerous significant strengths. The utilization of readily available clinical and laboratory indicators improves the practicality and scalability of the predictive model in resource-constrained environments like Egypt. Utilizing both statistical and machine-learning methodologies, we achieved a compromise between interpretability and strong prediction efficacy. The explicit presentation of model discrimination (ROC/PR curves) and calibration (Brier scores) enhances clinical trust in the results, while anchoring findings in effect-size estimates and feature relevance rankings fosters transparency and reproducibility. However, certain restrictions must be recognized. The cohort originated from a singular Egyptian Centre, exhibiting a prevalence of genotype 4; hence, external validation in broader and more diversified populations is necessary to ascertain generalizability. Only two SNPs were examined; more extensive genotyping panels or polygenic risk scores could yield further insights. The lack of longitudinal data and on-treatment markers, such as early viral kinetics, limits the assessment of dynamic predictors. The limited sample size ultimately constrains the ability to identify rare or nuanced predictor interactions. Our results indicate that fibrosis stage is the primary risk factor, with standard laboratory markers (hemoglobin, albumin, AFP) serving as independent contributors to DAA non-response, whereas the assessed NME1 SNPs provide no additional predictive value, corroborating and expanding upon a comprehensive worldwide data base. Advanced fibrosis and cirrhosis have been proven to be the most significant factors influencing unsatisfactory outcomes both before to and beyond the DAA period. Abdel-Ghaffar et al. presented a thorough analysis of genotype-4, highlighting elevated failure risks associated with advanced fibrosis, especially in Egypt where GT-4 is prevalent, reflecting our effect sizes corresponding to METAVIR stage increments (2). Recent longitudinal and biomarker investigations demonstrate that baseline fibrosis burden influences both the likelihood of sustained virologic response (SVR) and post-SVR outcomes, reinforcing our focus on fibrosis-based risk stratification (19),(20)\RL. Extensive cohorts of Egyptian genotype-4 patients treated with sofosbuvir–daclatasvir indicated that reduced albumin levels, anemia, and elevated AFP were significantly correlated with treatment failure (21),(22). The findings align with our multivariable analysis, which identifies albumin (OR 2.44), hemoglobin (OR 0.72), and AFP (OR 0.86) as independent predictors of non-response. Previous research from the interferon period similarly associated AFP and fibrosis with adverse outcomes, highlighting AFP’s persistent function as a nonspecific marker of liver damage and tumorigenic risk (23) . Outside of Egypt, composite clinical indicators that integrate age, fibrosis, albumin, and AFP (such as the General Evaluation Score) have been proven to forecast adverse outcomes even post-SVR (24)\RL. While IL28B/IFNL4 polymorphisms were powerful predictors in the interferon era (25), multiple analyses now show their attenuated or context-specific utility under DAA therapy (26),(27). Our null results for NME1 SNPs (rs2302254, rs16949649) reflect this transition. While IFNL4 variations may exhibit limited relationships in some subgroups (e.g., cirrhotic), their predictive significance is considerably inferior to that of phenotypic markers, hence endorsing the shift towards integrated clinical-laboratory models rather than reliance on single-gene predictors (28)\RL. Numerous Egyptian and worldwide research have investigated circulating miRNAs, especially miR-122, as supplementary indicators for treatment response (29),(30). Although associations are evident, the predictive performance remains poor (AUC <0.80), and technical requirements constrain practicality for extensive implementation. Supplementary non-invasive indicators, including hyaluronic acid, have been suggested for assessing fibrosis regression during sustained virologic response (SVR) (31), Enhancing our focus on fibrosis as a pivotal factor influencing baseline risk and long-term outcomes. Recent literature endorses multivariable and machine-learning methodologies for outcome prediction. In the HCV-TARGET registry, machine-learning methods (random forest, gradient boosting, neural networks) markedly surpassed conventional regression models in forecasting DAA failure (32). Likewise, Tada et al. demonstrated that integrated scoring systems that amalgamate age, fibrosis, and laboratory markers effectively predicted unfavourable liver outcomes even post-SVR. Comprehensive evaluations of AI-driven elimination frameworks underscore the significance of machine learning techniques in prioritising high-risk individuals and enhancing national hepatitis C virus elimination programs (28)\RL. Potential Egyptian cohorts consistently demonstrate elevated SVR rates (>90%) with sofosbuvir-based regimens, however identify the same principal predictors of non-response as observed in our study: age, fibrosis, albumin, alpha-fetoprotein, and platelets (21),(22),(33). Post-SVR follow-up investigations in Egypt indicate metabolic and oncogenic alterations despite achieving virological cure (34),(35), Highlighting the significance of pretreatment risk classification to identify patients necessitating enhanced surveillance despite attaining SVR. Collectively, these studies substantiate three conclusions that align with our findings: fibrosis establishes baseline risk and should be pivotal in treatment prioritisation; standard laboratory indicators such as albumin, haemoglobin, and AFP yield supplementary and clinically relevant predictive value; and single-variant pharmacogenetics provides minimal additional advantage in the DAA era relative to comprehensive phenotypic models. Our research enhances this body of literature by measuring the combined predictive efficacy of these markers in an Egyptian GT-4 cohort and providing calibrated, high-AUC models prepared for external validation and clinical application. Subsequent research must emphasise the external validation of this model inside bigger and more heterogeneous genotype-4 populations, as well as in cohorts comprising other HCV genotypes, to ascertain its generalisability. Integrating on-treatment dynamics, including quick virologic response and changing laboratory trends, could further refine risk assessments and improve predictive precision. Expanding genetic coverage by polygenic risk scores or comprehensive genomic panels also offers potential for identifying additional host factors influencing treatment outcomes. Ultimately, empirical implementation studies—encompassing cost-effectiveness assessments and preliminary integration into clinical workflows through electronic health records or mobile applications—will be essential to evaluate scalability, usability, and sustainability in standard practice. This study introduces a precise, scalable approach for forecasting DAA non-response in HCV genotype-4, utilizing commonly gathered data. Fibrosis, hemoglobin, albumin, and AFP are strong predictors; the examined SNPs are not. The model exhibits outstanding efficacy and provides a viable approach to personalized HCV treatment in Egypt and comparable environments. Subsequent validation and implementation studies will be essential for converting these findings into enhanced patient outcomes. Conclusion This work introduces the introductory integrated multi-biomarker prediction model for anticipating treatment non-response in Egyptian patients with HCV genotype 4 undergoing direct-acting antiviral therapy. By integrating clinical, hematological, biochemical, and genetic factors into a cohesive framework, we exhibited enhanced discriminative performance relative to models dependent on any singular domain. Our analyses determined that fibrosis stage is the most robust independent predictor of non-response, although standard laboratory markers, including hemoglobin, serum albumin, and AFP, also proved to be significant factors. Conversely, the assessed NME1 polymorphisms (rs2302254 and rs16949649), while biologically plausible, did not improve prediction accuracy, highlighting the restricted incremental use of candidate SNPs in the DAA era\RL. This work’s contribution is its methodological innovation and clinical relevance. The model quantifies the additive predictive value of normal laboratory data while demonstrating good calibration and discrimination, rendering it both scientifically robust and easily applicable in practical environments. This work offers a useful risk classification tool that enhances precision medicine strategies to optimize first-line cure rates and minimize retreatment expenses. These findings significantly enhance the field by illustrating that multi-domain, phenotype-centered models surpass single-marker methods, and they underscore the essential influence of fibrosis and hematologic reserve on treatment outcomes in genotype 4 populations. References 1. Organization WH. Hepatitis C [Internet]. Geneva: World Health Organization; 2025. Available from: https://www.who.int/news-room/fact-sheets/detail/hepatitis-c2. Abdel-Ghaffar TY, Sira MM, El Naghi S. Hepatitis C genotype 4: the past, present, and future. World J Hepatol. 2015;7(28):2792. 3. El-Ghitany EM. Hepatitis C virus infection in Egypt: current situation and future perspective. J high Inst public Heal. 2019;49(1):1–9. 4. Pawlotsky JM. Hepatitis C virus resistance to direct-acting antiviral drugs in interferon-free regimens. Gastroenterology [Internet]. 2016;151(1):70–86. Available from: https://pubmed.ncbi.nlm.nih.gov/27080301/5. Wyles DL, Dakappagari N, Naeger LK. Understanding hepatitis C virus drug resistance: clinical implications for current and future regimens. Clin Liver Dis [Internet]. 2017;21(3):603–17. Available from: https://pubmed.ncbi.nlm.nih.gov/28689563/6. Asselah T, Boyer N, Saadoun D, Martinot-Peignoux M, Marcellin P. Direct-acting antivirals for the treatment of hepatitis C virus infection: optimizing current IFN-free treatment and future perspectives. Liver Int [Internet]. 2016;36(Suppl 1):47–57. Available from: https://pubmed.ncbi.nlm.nih.gov/26725897/7. Ge D, Fellay J, Thompson AJ, Simon JS, Shianna K V, Urban TJ, et al. Genetic variation in IL28B predicts hepatitis C treatment-induced viral clearance. Nature [Internet]. 2009;461(7262):399–401. Available from: https://pubmed.ncbi.nlm.nih.gov/19684573/8. Chekuri S, Nickerson J, Bichoupan K, Sefcik R, Doobay R, Chang S, et al. Liver stiffness and biochemical response predict long-term liver-related outcomes in patients with chronic hepatitis C. Clin Gastroenterol Hepatol [Internet]. 2016;14(12):1788–97. Available from: https://pubmed.ncbi.nlm.nih.gov/27457432/9. Abdelwahab KS, Said ZNA. Hepatitis C genotype 4: What we know and what we don’t yet know. Hepatol Int [Internet]. 2016;10(1):38–44. Available from: https://pubmed.ncbi.nlm.nih.gov/26666504/10. Rizk HH, Hamdy NM, Al-Ansari NL, El-Mesallamy HO. Pretreatment predictors of response to PegIFN-RBV therapy in Egyptian patients with HCV genotype 4. PLoS One. 2016;11(4):e0153895. 11. AbdElrahman M, Ibrahim MK, Tawfik S, Omran D, Bendary MM, Hassanin SO, et al. The relation between SNPs in the NME1 gene and response to sofosbuvir in Egyptian patients with chronic HCV. Beni-Suef Univ J Basic Appl Sci. 2023;12(1):2. 12. Elabd NS, Tayel SI, Elhamouly MS, Hassanein SA, Kamaleldeen SM, Ahmed FE, et al. Evaluation of MicroRNA-122 as a biomarker for chronic hepatitis C infection and as a predictor for treatment response to direct-acting antivirals. Hepatic Med Evid Res. 2021;9–23. 13. Hoteit M, Mattar E, Allaw R, Abou Rached A. Epidemiological study assessing the overuse of proton pump inhibitors in Lebanese population. Middle East J Dig Dis. 2020;12(4):265. 14. Messina JP, Humphreys I, Flaxman A. Hepatitis C virus genotype 4: The past, present, and future. Hepatology [Internet]. 2015;61(1):77–87. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC4670951/15. Gomaa A, Gomaa M, Allam N, Waked I. Hepatitis C elimination in Egypt: story of success. Pathogens. 2024;13(8):681. 16. Abdel-Wahab R, Hassan MM, El-Ebeidy G. Genetic profile of Egyptian hepatocellular carcinoma patients associated with HCV genotype 4. BMC Res Notes [Internet]. 2008;1:106. Available from: https://bmcresnotes.biomedcentral.com/articles/10.1186/1756-0500-1-10617. Zekri ARN, Hafez MM, Bahnassy AA, Hassan ZK, Mansour T, Kamal MM, et al. Genetic profile of Egyptian hepatocellular-carcinoma associated with hepatitis C virus Genotype 4 by 15 K cDNA microarray: preliminary study. BMC Res Notes. 2008;1(1):106. 18. Pourhoseingholi MA, Vahedi M, Rahimzadeh M. Sample size calculation in medical studies. Gastroenterol Hepatol from bed to bench. 2013;6(1):14. 19. Rewisha E. Hyaluronic acid as non-invasive marker of fibrosis regression post-DAAs. Egypt Liver J [Internet]. 2021;11:24. Available from: https://pubmed.ncbi.nlm.nih.gov/34394258/20. Salama M. Long-term outcomes after SVR in cirrhosis patients. Hepatol Int [Internet]. 2024;18:1023. Available from: https://pubmed.ncbi.nlm.nih.gov/38920367/21. Ahmed OA, Safwat E. Sofosbuvir plus daclatasvir in treatment of Egyptian genotype 4 patients. Infect Disord Drug Targets [Internet]. 2018;18(2):137–43. Available from: https://pubmed.ncbi.nlm.nih.gov/29485230/22. Abdel-Moneim A. Real-life effectiveness and predictors of sofosbuvir-based regimens in genotype 4. J Med Virol [Internet]. 2018;90(10):1664–70. Available from: https://pubmed.ncbi.nlm.nih.gov/29900536/23. Al-Ashgar HI. AFP as predictor in interferon therapy. Saudi J Gastroenterol [Internet]. 2009;15(4):258–64. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC2813612/24. Tada T, Kumada T. General Evaluation Score predicts outcomes post-SVR. Sci Rep [Internet]. 2022;12:13740. Available from: https://pubmed.ncbi.nlm.nih.gov/35944354/25. Kanda T. IL28B in the DAA era: diminished relevance of host genetics. Int J Mol Sci [Internet]. 2015;16(12):30025–44. Available from: https://pubmed.ncbi.nlm.nih.gov/26690412/26. Ramamurthy N. IFNL4 variants and SVR in cirrhosis patients under DAAs. Liver Int [Internet]. 2018;38(10):1864–72. Available from: https://pubmed.ncbi.nlm.nih.gov/29754468/27. Pedergnana V. Genetic variation in IFNL4 locus across multiple genotypes. Hepatology [Internet]. 2019;69(2):630–43. Available from: https://pubmed.ncbi.nlm.nih.gov/30378188/28. Hur MH. Toward hepatitis C elimination using AI and machine learning. BMC Med Inf Decis Mak [Internet]. 2024;24(1):11. Available from: https://pubmed.ncbi.nlm.nih.gov/38232277/29. Elabd NS. Circulating miR-122 as predictor of DAA response in Egyptian genotype 4. Clin Exp Hepatol [Internet]. 2021;7(3):251–7. Available from: https://pubmed.ncbi.nlm.nih.gov/34660015/30. Hassuna NA. Circulating microRNAs and treatment response in HCV genotype 4. Egypt J Med Hum Genet [Internet]. 2022;23:20. Available from: https://pubmed.ncbi.nlm.nih.gov/35265321/31. Bojanic K, Bogojevic MS, Vukadin S, Sikora R, Ivanac G, Lucic NR, et al. Noninvasive fibrosis assessment in chronic hepatitis C infection: an update. J Clin Transl Hepatol. 2023;11(5):1228. 32. Park H, Lo‐Ciganic W, Huang J, Wu Y, Henry L, Peter J, et al. Machine learning algorithms for predicting direct‐acting antiviral treatment failure in chronic hepatitis C: An HCV‐TARGET analysis. Hepatology. 2022;76(2):483–91. 33. Abd El-Wahab EW. Long-term outcomes of generic DAAs in Egypt. Clin Liver Dis [Internet]. 2022;19(3):117–24. Available from: https://pubmed.ncbi.nlm.nih.gov/35850516/34. El-Ghandour A. Metabolic changes post-SVR in Egyptian HCV patients. Front Med [Internet]. 2023;10:11154. Available from: https://pubmed.ncbi.nlm.nih.gov/37065147/35. Salama M. Oncologic outcomes after DAA cure. Hepatol Int [Internet]. 2024;18:1023–32. Available from: https://pubmed.ncbi.nlm.nih.gov/38920367/ Information & Authors Information Version history V1 Version 1 24 September 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords antiviral agents epidemiology hepatitis c virus virus classification Authors Affiliations Mohamed AbdElrahman 0000-0001-5798-9671 [email protected] Al-Mustaqbal University College of Pharmacy View all articles by this author Marwa Khalil Ibrahim National Research Centre Biotechnology Research Institute View all articles by this author Ghadah Ali Al-Oudah Al-Mustaqbal University College of Pharmacy View all articles by this author Heba Salem F Beni Suef University Faculty of Pharmacy View all articles by this author Hossam Abuahmed Helwan University Faculty of Medicine View all articles by this author Metrics & Citations Metrics Article Usage 188 views 122 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Mohamed AbdElrahman, Marwa Khalil Ibrahim, Ghadah Ali Al-Oudah, et al. A Multi-Biomarker Prediction Model Integrating Genetics, Hematological Indices, and Fibrosis Stage to Forecast Non-Response in HCV Genotype 4. Authorea . 24 September 2025. DOI: https://doi.org/10.22541/au.175868487.76800218/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175868487.76800218/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fff93898d0109d6',t:'MTc3OTQ5MjI4Ng=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();