Optimizing spin‑column elution parameters for efficient HPV DNA recovery from preservative liquid‑based cytology cervical scrape samples

Introduction

Molecular biomarkers are rapidly used in clinical practice for disease detection, particularly in oncology, where they often outperform traditional diagnostic methods (for instance: Biopsy examination and visual inspection with acetic acid in cervical cancer screening) in sensitivity and specificity (1). Generating reliable molecular biomarker data requires high-quality clinical specimens, which can be challenging to obtain, particularly in low- and middle-income countries (2,3). Persistent barriers include limited numbers of trained personnel capable of performing proper sample collection, as well as logistical constraints related to transportation, cold-chain maintenance and storage (4). These challenges may compromise specimen integrity, ultimately affecting the accuracy and reliability of downstream molecular analyses.

Liquid-based cytology clinical samples, such as cervical scrapes collected with a brush and preserved in ThinPrep PreservCyt® solution, are widely used in human papillomavirus (HPV) detection for cervical cancer triage screening, as they efficiently preserve cellular material for both cytological and molecular analyses. However, a frequent bottleneck in downstream molecular assays, such as in viral load quantification, genotyping and sequencing, is inefficient DNA extraction, leading to a low yield or suboptimal purity of DNA (5,6).

To address this critical initial step, silica spin-column based extraction kits are widely used in both clinical and research laboratories due to their convenience, scalability and relative consistency. However, these standard protocols may not be optimal for all sample types or downstream applications, a limitation that is particularly acute for low-viral-load HPV samples. Therefore, the selection of elution parameters (particularly volume, incubation duration and the number of elutions) becomes a key optimization step, as they can significantly affect both the final DNA yield and the concentration of the eluate (7).

The optimization of protocol variables, such as incubation time, elution volume and number of elutions can significantly influence DNA yield and quality (8,9). Given that ThinPrep-preserved samples contain preservative agents, variable cell concentrations and potential inhibitors, protocol optimization is particularly relevant. Optimization research reveals trade-offs: Reduced elution volume increases the concentration, but may leave residual DNA on the column, while extended incubation improves recovery (10).

To the best of our knowledge, published data on the optimization of elution variables, specifically for the Monarch gDNA Spin Column kit (New England Biolabs, Inc.) using cervical swab clinical samples is relatively limited. Therefore, the Monarch kit was consciously selected due to its technical advantages in fundamental research, providing superior DNA purity, highly concentrated output and specialized protocols, such as HMW DNA protocols for downstream applications (11,12). The present study aimed to evaluate the effects of the following parameters on DNA concentration and purity from HPV ThinPrep cervicovaginal-scrape samples using the Monarch gDNA Spin Column kit: Incubation duration, elution volume and the number of elutions.

Materials and methods

Clinical specimen samples

A total of 17 cervical swab specimens were collected from women who underwent independent cervical cancer screening at the researchers' affiliated hospitals in Jakarta, Indonesia, including RSAB Harapan Kita (the author KH was originally affiliated with this hospital and changed affiliations during the study period as part of her professional appointment) and Gatot Soebroto Army Hospital (Indonesia Army Hospital), between August, 2024 and November, 2025. All specimens were processed and analyzed under the approvals granted by the Institutional Review Boards of both participating institutions. Samples were collected at the participating hospitals and subsequently transferred to the designated research laboratory for molecular analyses in accordance with the approved ethical protocols. Written informed consent was obtained from all participants. The cervical scrape samples were preserved in ThinPrep® PreservCyt® Solution (Hologic Inc.) and transported to the centralized main storage facility. Upon arrival, the samples were stored under controlled conditions at 4˚C, -20˚C or -80˚C until nucleic acid extraction. A total of 11 samples were initially maintained at 4˚C for up to 10 months, after which they were transferred to -20˚C for an additional 4 months of storage. In addition, 6 samples were stored at -80˚C for periods ranging from 2 to 10 months. Although the extended storage of PreservCyt aliquots at 4˚C exceeds the manufacturer's (Hologic Inc.) initial 21-day recommendations, previous large-scale evaluations have demonstrated that human genomic DNA remains structurally intact and viable for molecular amplification in PreservCyt at 4˚C for up to 2.5 years, provided that robust, crosslink-reversing extraction methodologies are employed (13). All samples were processed within the predefined acceptable storage timeframe prior to extraction. These samples were then aliquoted to ensure uniformity. The aliquots were used in parallel for the different extraction condition arms.

DNA extractions, quantification and purity assessment

Total genomic DNA was extracted using the Monarch Spin gDNA Extraction kit (cat. no. T3010; New England BioLabs, Inc.) as per the manufacturer's standard protocol through the binding and washing steps. This specific extraction system was selected for its optimized high-temperature lysis chemistry, which is required to effectively reverse the protein-nucleic acid crosslinking inherent to alcohol-based cytology media like PreservCyt (13). The extraction system was selected to facilitate efficient DNA recovery from preserved cervical specimens (14). Previous research has demonstrated that the optimization of digestion and incubation conditions can improve DNA recovery from fixed clinical samples (14). In the present study, extraction performance was evaluated based on DNA yield and purity, while DNA integrity was assessed using TapeStation analysis using a 4200 TapeStation System (Agilent Technologies, Inc.), as a quality control (15).

Samples were initially incubated at room temperature for 2 h. For the lysis step, following the addition of lysis buffer, Proteinase K and RNAse A which is included with the extraction kit Monarch gDNA Spin Column kit (New England Biolabs, Inc.) were added to the samples which are in 1.5 ml-microcentrifuge tube. For optimized lysis conditions, samples were incubated under at 56-60˚C for the specified duration before proceeding to subsequent binding steps; for the two-elution condition, eluates were collected either as separate fraction (to quantify incremental recovery) or pooled when total DNA mass was the endpoint. To determine the optimal conditions for reversing PreservCyt crosslinking and maximizing target recovery, three parameters were systematically evaluated: lysis incubation duration at 56-60˚C (30, 40, 50, 60 and 80 min), elution volume (30, 50, 60 and 75 µl), and the number of elution cycles (single vs. sequential double elution using the provided elution buffer).

Due to limited specimen volume, each treatment condition (incubation time x elution volume x elution count) was performed using replicate aliquots derived from a single homogenized specimen per patient. For each patient, 4 ml of sample was homogenized and aliquots were randomly assigned to extraction conditions to reduce potential batch and processing biases during DNA extraction. Eluted DNA concentration was quantified using a fluorescence-based assay on the Quantus™ Fluorometer CAT E6150 (Promega Corporation), while purity was assessed by spectrophotometry (NanoDrop) using A260/A280 and A260/A230 ratios; samples were classified as ‘pure’ when A260/A280 was 1.8-2.0 and A260/A230 was ≥2.0(16), and samples outside these thresholds were categorized as potentially affected by contaminants.

DNA integrity analysis

DNA quality and quantity were evaluated as part of the quality control assessment for the extraction optimization workflow using the 4200 TapeStation system (Agilent Technologies, Inc.). Samples were prepared following the manufacturer's protocols, utilizing the Genomic DNA ScreenTape assay (cat. no. 5067-5365). Data were analyzed using Agilent TapeStation software version 5.1 (Agilent Technologies, Inc.).

DNA integrity was evaluated by determining the distribution of fragment sizes, expressed as the percentage of DNA within specific size regions relative to the total analyzed DNA. The upper and lower size thresholds for quantification were defined based on the fragment size profile within the instrument's measurement range, with a lower threshold set at 100 bp. The DNA integrity number (DIN) was calculated based on the proportion of DNA fragments above the lower limit within the functional DIN range (17).

Experimental design

The primary response variables evaluated in the present study were DNA yield (ng/µl) and purity status, categorized as either ‘pure’ or ‘contaminated’. The experimental parameters were defined as the independent variables, namely: Incubation duration (t), elution volume (v) and elution cycle (e). The objective of this analysis was to identify the optimal combination of these parameters to maximize both yield and purity levels.

To evaluate the influence of varying experimental conditions within the constraints of limited specimen volumes, a fractional factorial design was employed. This approach utilized specific combinations of the previously defined incubation durations and elution volumes, resulting in a total of nine distinct treatment configurations for ‘Yield’ and ‘Purity’ (Table SI). This specific layout was necessitated by the high value and limited availability of the 17 clinical specimens. In clinical optimization studies, the \Sparsity of Effects\ principle suggests that primary system behavior is driven predominantly by main effects and low-order interactions. By selecting cells within the 16-cell grid (focusing on the 40-min and 80-min duration tiers), the characterization of the most promising operational ranges was prioritized, while maintaining ethical and practical stewardship of patient samples. This approach allowed for the exploration of the experimental scope more broadly than a narrow, balanced design of fewer parameters would have permitted.

Statistical analyses

The statistical analyses in the present study were designed to achieve two primary objectives. First, mixed-effect models were employed to evaluate the association between the response variables and the independent parameters. This initial step served to validate the hypothesis that DNA ‘Yield’ and ‘Purity’ are influenced by the three selected aforementioned parameters. Notably, the focus of this specific analysis was to establish the existence of these associations (i.e., providing evidence of statistical significance) rather than quantifying the magnitude of the effect sizes. To account for variability in storage conditions, this factor was included as a random effect in all mixed models. Second, response surface methodology (RSM) was employed to identify the optimal parameter settings for each individual response variable and, subsequently, to determine the global optimum for both responses simultaneously.

Linear mixed-effect model for ‘Yield’. Linear mixed-effects models were employed to estimate the fixed effects of incubation duration (t), elution volume (v), and the number of elutions (e) on the continuous outcome variables: DNA yield (ng/µl). The mixed-effect modeling approach was selected to account for the hierarchical data structure in which multiple technical replicates (k) are nested within each treatment combination (j) applied to the same biological sample (i), thereby requiring mixed-effects to accommodate within-sample correlation (18).

The experimental parameters (t, v, e) were included as fixed effects with a full factorial structure encompassing all two- and three-way interactions. Between-sample biological variation was modeled through a by-sample random intercept bi, assumed to be normally distributed with zero mean and variance . The model specification for response yijk (the k-the replicate from the i-th sample under the j-th treatment combination) is:

where the global intercept (β0) represents the baseline response. Coefficients β1 to β7 quantify the fixed effects of predictors and their interactions. The random intercept bi captures sample-specific deviations from the global mean, assumed normally distributed with variance . Residual error εijk accounts for unexplained within sample variability, following a normal distribution with variance σ2.

Observations with missing values were excluded listwise. Model parameters were estimated via restricted maximum likelihood (REML) using the lme4 package (version 1.1-35.1) (19). The significance of fixed-effect coefficients was assessed using t-tests with Satterthwaite's degrees-of-freedom approximation, implemented in the lmerTest package (20).

Generalized linear mixed-effect model for DNA ‘Purity’. To analyze the factors influencing extraction purity, a binary composite response variable was constructed based on the previously defined spectrophotometric ratios, categorized as ‘Pure’ and ‘Contaminated’. This binary response was analyzed using a Generalized linear mixed model (GLMM) with a logit link function. This framework is particularly suited for clustered experimental outcomes and non-Gaussian responses, as it explicitly accounts for the nested nature of laboratory measurements (21). Wang et al (22), introduced a mixed-effects logistic regression framework designed for large genomic case-control studies, illustrating the capacity of GLMMs to accommodate correlated observations and heterogeneity across experimental units. Similarly, Noma and Gosho (23), discussed the utilization of logistic mixed-effects models for clustered binary clinical outcomes to emphasize the importance of random effects in accounting for within-subject and within-group dependence.

Let denotes the purity outcome for the k-th observation from the i-th biological sample under the j-th treatment combination, with . The model is specified as:

;

.

The linear predictor included the same fixed-effect factors and interactions as the continuous model, plus a random intercept for biological sample:

where γ0 is the fixed global intercept on the log-odds scale, γ1, …, γ7 are the fixed-effect coefficients for the main effects and interactions, and is the random intercept for sample.

The model was fitted via maximum likelihood estimation using the glmer function from the lme4 package in R (19), with a binomial family specified. Wald z-tests were used to evaluate the significance of the fixed effects. The model was fitted via maximum likelihood estimation using the glmer function from the lme4 package in R (19), with a binomial family specified.

RSM optimization

Following the inferential analyses, RSM was employed to visualize the mixed models on each DNA ‘Yield’ and ‘Purity’ responses. The response surfaces were graphically represented using two-dimensional contour plots and three-dimensional surface plots. Stationary points within the fitted models were determined through canonical analysis, adhering to established response surface methodology diagnostic protocols (24). Multi-response optimization was carried out by jointly considering the predicted outcomes for yield and purity over the entire design space.

Results

Mixed model analyses revealed that the evaluated experimental factors accounted for a small proportion of the variance in each response, as indicated by the R2 values shown in Table I. This suggests that other unobserved factors substantially contribute to the overall variability. A notable difference in the modeling results was observed between the LMM for DNA yield and the GLMM for DNA purity. Based on the standardized coefficients and model selection using AIC and likelihood ratio tests (LRT), the removal of individual predictors from the full model did not significantly worsen model fit, as all LRT P-values exceeded the conventional threshold of 0.05. However, among the yield predictors, the number of elutions (e) exhibited the largest change in AIC when removed, suggesting its relatively stronger contribution, while the interaction term between incubation duration (t) and elution volume (v) contributed minimally (Table I).

Table I Summary of the results of mixed models on ‘Yield’ and ‘Purity’ responses.

Table I

Summary of the results of mixed models on ‘Yield’ and ‘Purity’ responses.

Response variable	Fixed-effect variable	Std. coefficient	P-value	AIC	P-value (LRT)	R2
Yield	Incubation duration (t)	0.096	0.283	557.627	0.216	R²m=0.242, R²c=0.479
	Elution volume (v)	-0.061	0.587	556.335	0.579	R²m=0.266, R²c=0.467
	No. of elutions (e)	-0.262	0.174	559.177	0.137	R²m=0.239, R²c=0.434
	Incubation duration² (t²)	-0.04	0.586	555.468	0.528	R²m=0.260, R²c=0.482
	Elution volume² (v²)	-0.072	0.167	556.462	0.141	R²m=0.244, R²c=0.418
	Incubation duration x elution volume (t.v)	-0.021	0.912	557.121	0.904	R²m=0.267, R²c=0.482
Purity	Incubation duration (t)	-0.615	0.538	47.748	0.776	R²m=0.181, R²c=0.483
	Elution volume (v)	-0.539	0.59	47.734	0.796	R²m=0.175, R²c=0.485
	No. of elutions (e)	1.426	0.154	49.857	0.139	R²m=0.098, R²c=0.418
	Incubation duration² (t²)	-1.133	0.257	48.652	0.321	R²m=0.129, R²c=0.453
	Elution volume² (v²)	-0.877	0.381	49.244	0.209	R²m=0.144, R²c=0.455
	Incubation duration x elution volume (t.v)	1.613	0.107	48.257	0.443	R²m=0.157, R²c=0.441

[i] Standardized coefficients are shown for Yield. P-values for Yield were obtained using Wald t-tests with Satterthwaite's degrees-of-freedom approximation, while P-values for Purity were obtained using Wald z-tests. AIC and R² values are from reduced models with the corresponding term removed. R²m=marginal R² (fixed effects only); R²c=conditional R² (fixed + random effects). Storage temperature was included as a random effect.

For DNA purity, the number of elutions exhibited the largest standardized coefficient, indicating a potential role in explaining purity, although its contribution was not statistically significant. The R2 values further indicated that fixed effects explained only a small to moderate proportion of the variance, with marginal R2 values ranging from 0.098 to 0.0267, while the inclusion of random effects substantially increased the explained variance, with conditional R2 values ranging from 0.418 to 0.485 (Table I). This suggests that unmeasured factors beyond the experimental parameters evaluated in the present study may play a substantial role in determining both DNA yield and purity.

Jackknife resampling analyses were performed to evaluate the stability of both mixed-effects models, with detailed resampling results presented in Tables SII and SIII. These analyses demonstrated high agreement with the models derived from the full dataset; specifically, 17 out of 17 resampled models for DNA yield, and 10 out of 17 models for DNA purity, were consistent with the primary full models. These analyses demonstrate that despite the limited sample size and the use of second-order modeling, the results remain robust and stable.

Subsequently, the previously described mixed-effects models were mapped using RSM. It is important to emphasize that, herein, RSM was utilized strictly as an analytical tool to visualize and extract optimal points from the existing second-order mixed models, rather than to fit new independent models. Both response variables were first optimized individually based on the two continuous parameters using their corresponding mixed-effects models. For DNA yield, the RSM analysis identified a stationary optimum at 77 min of incubation duration and an elution volume of 42 µl, producing a predicted yield of 4,420 ng/µl. Although the mixed model for DNA purity lacked statistically significant predictors, its optimum parameters were still assessed via RSM visualization. The optimum point for DNA purity was achieved at 80 min of incubation and an elution volume of 75 µl, resulting in a predicted purity probability of 0.99, indicating a very high likelihood of achieving acceptable DNA purity under these conditions (Fig. 1).

Figure 1 Visualization of response surface methodology results. Columns represent the visualization type: 2D contour plots (left panels) and 3D response surface plots (right panels). Rows represent the optimized parameters: (A) ‘Yield’ and (B) ‘Purity’. The green dotted lines indicate the predicted optimum values for each response variable.

Identifying experimental parameters that satisfy both responses simultaneously, the mean values of the individual optima were calculated and presented in Table II, resulting in a ‘compromised’ condition of a 59-µl elution volume and 79 min of incubation. This averaged parameter set represents a shift from the yield optimum, specifically an increase of 17 µl in elution volume and an increase of 2 min in incubation time, as well as a shift from the purity optimum, namely a decrease of 17 µl in elution volume and a decrease of 1 min in incubation time, thereby providing a balanced trade-off between DNA yield and purity (Fig. 1). Providing a quantitative basis for the optimization, the final fitted second-order polynomial equations for DNA yield and purity obtained from the mixed-effects RSM method are given below:

Table II Optimization results based on the mixed models for the response variables.

Table II

Optimization results based on the mixed models for the response variables.

Optimization approach	Response	Volume (v)	Duration (t)	Prediction response and its 95% confidence interval (CI)
Single-response	Yield	42	77	5,247.567 (859.110; 9,503.83)
	Purity	75	80	0.999 (0.177; 1.000)
Multi-responses	Yield	59	79	4433.583 (-142; 9099.336)
	Purity			0.966 (0.154, 0.999)

[i] Optimization approach: Single-response optimization refers to the determination of optimal conditions for each response variable independently based on its respective RSM model. Multi-response optimization was applied to satisfy both responses simultaneously by calculating the mean values of the individual optimal settings, resulting in a compromised dual-response optimum. Confidence intervals: Values in parentheses represent the 95% confidence intervals (CI) for the predicted response values at the corresponding optimum points. Elution condition: All predictions for DNA yield and purity were generated using Elution = 1, which was identified as the optimal elution condition based on model estimates.

DNA integrity quality control

The quality assessment of DNA extracted from HPV samples (001, 002, 003, 004, 005A, 005B, 005C, 006, 007, 009, 010, 011, 014, 015, 016, 017, 018 and 019) using the TapeStation platform demonstrated considerable variability in DNA integrity and fragment size distribution among samples. Based on the DIN values presented in Table SIV, samples HPV 002, HPV 003, HPV 004, HPV 005A, HPV 014, HPV 017 and HPV 018 exhibited relatively high DNA integrity (DIN ≥7.0). Among these, HPV 003 showed the highest DIN value (8.4), followed by HPV 014 and HPV 017 (DIN 7.9), indicating excellent preservation of genomic DNA. HPV 002 and HPV 018 also demonstrated high integrity with DIN values of 7.6 and 7.0, respectively. By contrast, HPV 001 exhibited a comparatively low DIN value (4.5), suggesting substantial DNA degradation. Samples HPV 007, HPV 011, HPV 015, HPV 016 and HPV 019 displayed moderate DNA integrity, with DIN values ranging from 5.1 to 6.4, indicating partial fragmentation of genomic DNA.

Notably, the DNA concentration was not consistently associated with DNA integrity. Several samples with a high DNA yield, including HPV 019 (8,150 ng), HPV 016 (7,075 ng) and HPV 018 (8,400 and 7,300 ng), did not necessarily exhibit the highest DIN values or the most intact fragment profiles. By contrast, samples HPV 003, HPV 004 and HPV 014, which exhibited comparatively moderate DNA concentrations, demonstrated superior DNA integrity characterized by higher DIN values and dominant HMW DNA fragments. These findings indicate that elevated DNA concentration alone is insufficient to predict DNA quality, as high DNA yield may include fragmented or partially degraded DNA molecules. Overall, these results indicate that DNA integrity is more strongly associated with the preservation of long DNA fragments and elevated DIN values than with total DNA concentration alone. Samples demonstrating a high DNA integrity, particularly HPV 003, HPV 004, HPV 014, HPV 017 and HPV 018, may therefore be more suitable for downstream long-read sequencing applications. By contrast, samples with moderate DNA integrity may remain appropriate for amplification-based molecular assays, such as PCR or qPCR targeting short amplifications. Samples with a lower DNA integrity, including HPV 001, may require further optimization of extraction procedures and storage conditions to improve DNA preservation for advanced molecular analysis.

Discussion

The present study aimed to identify the optimal DNA extraction parameters (incubation duration (t), elution volume (v) and the number of elutions (e) and DNA integrity analysis from cervicovaginal specimens in ThinPrep buffer, a matrix often characterized by low cellularity and high dilution. The experimental data, albeit derived from a limited sample size, suggest that incubation time, elution volume and the number of elutions each contributed to extraction performance, with the number of elutions exhibiting the most consistent statistical association with DNA yield. These preliminary results may have implications for supporting more consistent DNA recovery in downstream molecular diagnostics. In clinical screening workflows, particularly for specimens with low cellularity, even minor variations in extraction efficiency can influence the success of downstream molecular detection. This is particularly relevant in resource-limited settings where specimen quality is often compromised, though further validation with larger sample cohorts is warranted.

The RSM fitted using the mixed-effects framework, evaluated the effects of incubation duration (t), elution volume (v) and the number of elution cycles (e) on DNA yield and purity. As the predicted optimum conditions for DNA yield and DNA purity occurred at different combinations of incubation time and elution volume, an operational compromise condition was determined by averaging the coordinates of the two individual optima. This approach resulted in a compromise condition of approximately 79 min of incubation and 59 µl of elution volume (Table II), providing a balanced operating condition that maintained satisfactory DNA yield while preserving DNA purity. The higher efficiency of the first elution observed in the present study is consistent with the desorption kinetics of silica membranes, where DNA release is greatest during the initial contact with the elution buffer and decreases in subsequent elution cycles. Although additional elution cycles may increase total DNA recovery the incremental gain becomes progressively smaller indicating diminishing returns (8).

Furthermore, the present study observed a trade-off between total yield and concentration governed by elution volume. Previous research has reported that low-volume elution can produce more concentrated nucleic acid eluates and may improve recovery of low-abundance targets (9). The data presented herein suggest that within the 30-80 µl range, these effects act independently, allowing for flexibility based on the specific requirements of downstream applications. Lower elution volumes which yield higher DNA concentrations are more advantageous for next-generation sequencing (NGS) workflows, whereas higher elution volumes producing more dilute eluates are better suited for PCR-based applications (25,26). This optimization aligns with a broad body of literature [e.g., Akahane et al (27); Shibata et al (6) and Naegele et al (5)] showing that the delicate balance between DNA quantity, purity and concentration is the primary determinant of success in sensitive molecular assays.

Consequently, for laboratories processing ThinPrep specimens with silica-spin kits, prioritizing an extended incubation of ~77 min over increasing elution volumes or adding redundant elution steps to maximize DNA yield may be recommended. For maximizing DNA purity, an extended incubation of ~80 min is recommended instead. While a second elution remains a viable strategy when maximum DNA mass is required for replicate assays, it should be noted that this approach provides diminishing returns and does not substantially alter purity metrics. Ultimately, elution volumes in the mid-range should be selected based primarily on the concentration requirements of the specific downstream workflow, such as NGS or qPCR, as this flexibility allows for standardized, cost-effective protocols that are compatible with existing laboratory infrastructure in resource-limited settings. By maintaining measurable DNA quality and integrity profiles suitable for molecular applications, these optimized extraction conditions may support standardized extraction workflows compatible with downstream molecular analyses.

The optimization results demonstrated that achieving high DNA purity required a relatively larger elution volume compared to conditions that maximized DNA yield or those recommended by the Monarch extraction protocol. This observation is likely influenced by both intrinsic sample quality and the physicochemical behavior of silica-based extraction columns. Low-biomass samples require higher elution volumes to improve DNA release from silica membranes, while insufficient volumes result in incomplete recovery and reduced yield and purity (28,29).

Previous research has reported that increasing elution volume enhances DNA desorption efficiency under low-salt conditions, particularly in silica column-based extraction systems, albeit at expense of DNA concentration (8). Higher elution volumes enhance DNA purity by minimizing co-elution of residual contaminants (28,30). During centrifugation, a larger elution volume can promote more effective spatial separation eluted DNA and contaminants retained within the silica membrane or column matrix, thereby minimizing carryover into the final eluate (29).

In the present study, the purity cut-off values applied for DNA quality assessment were not strictly based on the minimum input specifications recommended by the NGS library preparation kit, but were instead determined through a combination of empirical considerations and evidence from recent literature. Although manufacturers recommend A260/280 (1.8-2.0) and A260/230 (2.0-2.2) ranges, recent studies show long-read sequencing performance depends more on DNA integrity and fragment length than strict spectrophotometric thresholds (12,31). This approach aligns with recent reports suggesting that functional validation through sequencing performance provides a more robust assessment of DNA suitability than reliance on purity ratios alone (5,27).

Furthermore, the findings of the present study highlight that utilizing DNA integrity as a fundamental quality control step reveals a lack of consistent correlation with total DNA concentration. As a quality control indicator, high DIN values are highly preferred; previous research indicates that DIN strongly predicts the success of downstream analyses, especially NGS (32). On the other hand, quality control assessments demonstrating a predominance of short fragments reflect DNA degradation driven by biological and environmental processes, leading to progressive fragmentation (33).

The extraction process parameters, particularly lysis duration and temperature, play a decisive role in preserving long DNA fragments. These factors are particularly critical in PreservCyt samples, where protein-DNA crosslinking requires controlled lysis conditions to prevent additional fragmentation (34,35). In addition, the findings of the present study revealed cases of high DNA concentration that do not exhibit broad fragment size distribution, suggesting that DNA quantity does not necessarily reflect its functional quality. This condition is likely due to the accumulation of degraded DNA, emphasizing the importance of evaluating fragment size distribution alongside quantification (33,36).

Moreover, the observation that increasing elution volume and the number of elution cycles does not consistently improve DNA integrity highlights the limitations of conventional approaches that focus primarily on maximizing DNA recovery. Although yield optimization is often used as a key indicator of extraction success, particularly in limited clinical samples (37), the findings of the present study suggest that such strategies may overlook critical aspects of DNA quality. As previously reported by Simbolo et al (36) short DNA fragments may dominate even when the total yield is high. Therefore, the presnet study expands the current perspective by emphasizing that extraction optimization should balance both DNA quantity and integrity.

The observed variability in DNA quality also points to the influence of pre-analytical factors, such as sample storage and handling, which can accelerate DNA degradation and affect final analytical outcomes (38). In practice, this has a direct impact on the efficiency of downstream application, including sequencing performance. High-integrity DNA is essential for achieving optimal analytical results (35). The heterogeneous DNA integrity observed in the present study may be attributed to variations in storage conditions (4, -20 and -80˚C) and storage durations. These findings are consistent with and further extend those of previous reports indicating that DNA degradation is cumulative and strongly influenced by storage conditions and sample handling (33,38); in the other words, the stability reported in the literature does not entirely eliminate the risk of fragmentation if not accompanied by optimal extraction conditions.

Overall, DNA integrity assessment provides complementary quality-control information beyond yield and purity measurements and may assist in evaluating sample suitability for downstream molecular applications. The optimization analyses suggested that elution conditions were more strongly associated with DNA recovery than incubation duration or elution volume within the tested range. Collectively, the findings support the use of a single-elution workflow and provide a framework for optimizing DNA extraction from ThinPrep specimens, while acknowledging that further validation using downstream molecular assays is required (8). Collectively, these findings confirm that elution protocol primarily determines DNA yield while storage temperature serves as a confounding factor affecting both yield and purity. Based on these results, the recommended protocol for optimal HPV DNA extraction is using the elution 1 method, with an incubation time of 77 min and an elution volume of 41 µl to maximize yield, while storage at -80˚C is advised to maintain DNA purity. While a second elution remains a viable strategy when maximum DNA mass is required for replicate assays, it should be noted that this approach provides diminishing returns and does not substantially alter purity metrics. Ultimately, elution volumes in the mid-range should be selected based primarily on the concentration requirements of the specific downstream workflow, such as NGS or qPCR, as this flexibility allows for standardized, cost-effective protocols that are compatible with existing laboratory infrastructure in resource-limited settings (39,40). By maintaining measurable DNA quality and integrity profiles suitable for molecular applications, these optimized extraction conditions may support standardized extraction workflows compatible with downstream molecular analyses.

The present study provides a framework for optimization, several limitations remain that offer clear avenues for future research. First, while the present study focused on yield and purity, it did not directly measure downstream performance metrics, such as read length distribution, mapping rates, or HPV detection sensitivity. Establishing this direct link via qPCR Ct shifts and NGS coverage remains a critical next step to validate the clinical utility of these optimized parameters. Additionally, although spectrophotometric ratios (namely: A260/A280 and A260/A230) provide essential purity data, they do not fully account for the presence of PCR inhibitors. Future studies incorporating qPCR inhibition assays, spike-in controls, or the testing of supplemental techniques, such as elution buffer warming and the addition of carrier nucleic acids would provide more in-depth biochemical insight into enhancing recovery from samples with critically low cellularity. Finally, as laboratories move toward high-throughput diagnostic settings, a formal cost-benefit analysis will be essential for economic optimization. As noted by Neal et al (41), such low-cost optimizations are increasingly vital for preserving the high-quality DNA required for advanced molecular testing in resource-constrained environments, ensuring that procedural efficiency does not come at the expense of diagnostic accuracy.

In conclusion, the present study demonstrated that the optimization of pre-analytical extraction parameters using linear mixed models identified incubation time as a contributor to DNA yield variation from cervicovaginal samples in ThinPrep buffer without compromising purity. Incubation time and elution volume independently influenced DNA recovery, suggesting opportunities to optimize extraction conditions according to downstream concentration requirements. By prioritizing procedural optimization over increased buffer volume, this approach may improve recovery efficiency, particularly in low-cellularity samples. However, as downstream functional performance was not evaluated, these findings should be considered preliminary recommendations pending validation in diagnostic workflows. Overall, the results presented herein provide a potentially cost-effective strategy to improve DNA extraction efficiency and inform future studies on clinical and molecular performance.

Supplementary Material

Experimental design matrix and parameter levels for the fractional factorial design.

Jackknife resampling analysis for evaluating the stability of the linear mixed model for DNA yield.

Jackknife resampling analysis for evaluating the stability of the generalized linear mixed model for purity.

DNA integrity number from TapeStation, and quality assessment, storage temperature and DNA yield of HPV cervicovaginal ThinPrep samples.

Acknowledgements

The authors wish to express their gratitude to Professor Renske Steenbergen and Ms. Annina van Splunter (VU University Medical Center, Amsterdam, The Netherlands) for their insightful initial discussions regarding the optimization of DNA extraction. The authors would also like to thank Dr Ahmad R. Utomo and Mrs. Kinasih Prayuni (Yarsi University, Jakarta, Indonesia) for their valuable feedback on the optimization results and their suggestions for evaluating DNA quality. In addition, the authors would like to thank Mr. Abdul Azizul Hakim and Mrs. Christine Sembiring (Yayasan Satriabudi Dharma Setia, Tangerang, Indonesia) for their exceptional management of research logistics and for facilitating cross-institutional discussions.

Funding

Funding: The wet-lab experiments, subsequent laboratory analyses and data interpretations were supported by research funding from PaninBank through the Yayasan Satriabudi Dharma Setia initiative to advance genomic and molecular diagnostic research in Indonesia. Downstream data analyses and part of manuscript preparation were supported by Siena Clinical under Internal grant no. MCI-XI26-002.

Availability of data and materials

The data generated in the present study may be requested from the corresponding author.

Authors' contributions

NKI, NAF, HMARP, KH and PWN conceptualized the study. AKP conceptualized the laboratory experiments and experimental design. NKI and PWN were involved in the study methodology. NKI, WH, PS and NTGP were involved in the investigative aspects of the study. PRM contributed to sample collection. NKI, WH, PS, PRM and NTGP were involved in data curation. NKI, MAHPI, NAF, IW and PWN were involved in the formal analysis. MAHPI was involved in visualization. PRM and IW were involved in project administration. MAHPI and PWN were involved in the writing of the original draft of the manuscript. IW and PWN were involved in the writing, review and editing of the manuscript. VSWB, HMARP, KH and PWN supervised the study. VSWB and AKP was involved in data validation. VSWB and NKI confirm the authenticity of all the raw data presented in this study. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

The study was approved by the Medical Ethics Review Committee of the RSAB Harapan Kita (Jakarta Indonesia; approval no. IRB/04/02/ETIK/2023) and the Indonesia Army Hospital (Jakarta, Indonesia; approval no. 127/XI/KEPK/2024). Writen informed consent was obtained from all participants.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References