BOC Sciences provides physicochemical prediction services to help research teams understand how molecular structure and environmental conditions influence compound behavior. By combining property-specific computational models, molecular descriptors, analogue data, and experimental evidence when available, we support compound comparison, library prioritization, lead optimization, experimental planning, and preformulation research within broader computer-aided drug discovery programs.
Physicochemical prediction is the computational estimation of molecular properties—such as lipophilicity, ionization constants, aqueous solubility, permeability, and molecular descriptors—from chemical structure alone. Rather than relying solely on experimental measurement, which can be time-consuming and resource-intensive for large compound collections, computational prediction applies validated algorithms, QSPR models, quantum chemical calculations, and machine learning methods to generate property data rapidly and at scale. In drug discovery, these predictions inform compound prioritization, guide structure–property relationship analysis, flag potential developability issues early, and help medicinal chemists design molecules with balanced property profiles before committing to synthesis and assay resources.
BOC Sciences predicts octanol-water partition coefficients (LogP) and pH-dependent distribution coefficients (LogD) using multiple complementary approaches, including atom-based, fragment-based, and property-based algorithms, to deliver reliable lipophilicity estimates across diverse chemotypes.
We predict acid–base dissociation constants using Hammett-type linear free energy relationships, quantum chemical calculations, and machine learning models trained on curated experimental pKa datasets.
BOC Sciences estimates intrinsic and pH-dependent aqueous solubility using general solubility equations, fragment-based models, and machine learning approaches trained on curated thermodynamic and kinetic solubility datasets.
Our service calculates a comprehensive panel of 2D and 3D molecular descriptors and evaluates compounds against established drug-likeness and lead-likeness filters to support compound quality assessment.
BOC Sciences predicts passive membrane permeability and absorption-related properties using a combination of physicochemical descriptor-based models and in silico membrane-interaction simulations.
We predict chemical stability and reactivity properties that influence compound handling, storage, and experimental behavior, using quantum chemical calculations and structure-based reactivity models.
BOC Sciences helps research teams move from compound structures to ranked property profiles, identifying lipophilicity trends, solubility risks, and permeability bottlenecks across an analogue series so that synthesis and assay resources are focused on the most promising candidates.
BOC Sciences delivers customized physicochemical prediction support across a range of project scales and objectives—from focused single-compound profiling to large library screening campaigns. Key project categories include:
| Project Type | Service Scope & Key Outputs |
| Single-Compound Property Profiling | Comprehensive physicochemical characterization of an individual compound, covering LogP/LogD, pKa, solubility, permeability estimates, key molecular descriptors, and drug-likeness assessment. Delivered as a detailed property report with structure-annotated predictions and interpretation notes for decision-making. |
| Analogue-Series Structure–Property Comparison | Systematic property prediction across a congeneric series with matched molecular pair (MMP) analysis to isolate the impact of specific structural modifications on lipophilicity, solubility, and permeability. Outputs include property trend visualizations, structure–property relationship tables, and design recommendations for the next iteration. |
| Compound Library Screening and Ranking | High-throughput computational screening of compound libraries against multiple property filters—drug-likeness, lead-likeness, solubility, permeability, and reactivity alerts. Compounds are ranked and categorized by overall property profile quality, enabling efficient triage before experimental screening or synthesis. |
| Condition-Dependent Property Prediction | Prediction of how compound properties change under varying conditions: pH-dependent LogD and solubility profiles, temperature-dependent stability estimates, and ionic-strength effects. Supports formulation buffer selection, assay condition planning, and stability risk assessment. |
Share your compound structures (SMILES, SDF, or MOL format), the properties of interest, the pH and temperature conditions relevant to your assays, and any experimental reference data you have. Our computational scientists will design a project-specific plan covering model selection, consensus scoring across multiple prediction algorithms, QSAR prediction integration where applicable, and a complete property report with interpretation and design guidance.
BOC Sciences receives compound structures in standard formats (SMILES, SDF, MOL, or CDX), performs structure standardization, removes counterions, generates tautomers and protonation states at defined pH values, and confirms with the client the property panel, conditions, and any experimental reference data to be used for model benchmarking.
Our team selects the appropriate prediction algorithms for each property based on the compound's structural class, molecular weight range, ionizable group profile, and available experimental data. Where applicable, we apply consensus scoring from multiple orthogonal methods and flag predictions that fall outside the model's defined applicability domain.
BOC Sciences compiles predicted properties into structured reports with tabular summaries, property distribution plots, pH-dependent profile curves, and radar charts for multi-parameter comparison. Each prediction is accompanied by confidence metrics and, where relevant, literature reference values for structurally similar compounds to support interpretation.
Clients receive the complete property prediction report together with an interpretation summary highlighting key property risks, structure–property trends, and specific recommendations for structural modifications to improve compound profiles. Our team remains available for follow-up discussion and iterative prediction rounds as the chemical series evolves.
Partner with BOC Sciences for tailored prediction of LogP, LogD, pKa, aqueous solubility, permeability, stability, and other physicochemical properties. We select suitable computational models and provide clearly interpreted results to support compound comparison, experimental planning, and formulation research.
Structural features associated with low solubility, excessive lipophilicity, unfavorable ionization, weak permeability potential, or chemical instability can be identified before extensive testing. Early visibility helps teams investigate the most consequential risks while molecular changes are still practical.
Computational triage narrows large compound sets to representative, high-priority, or uncertain candidates. Experimental resources can then be directed toward model validation, boundary cases, and the measurements most likely to change a project decision.
Standardized calculations make it possible to compare many structures on the same property scale. Multi-parameter views reveal trade-offs that a single endpoint can miss and support transparent selection of compounds for synthesis or follow-up testing.
Predicted pKa, logD, solubility, polarity, and stability trends help define useful pH ranges, buffer systems, concentration windows, analytical methods, and formulation hypotheses. Predictions guide study design while experimental data remain the basis for confirmation.
Predicted LogD, solubility, and pKa values can shift significantly with pH, temperature, ionic strength, and buffer composition, yet many prediction tools provide only a single-condition estimate. BOC Sciences addresses this by generating property predictions across a client-defined condition matrix—for example, LogD at pH 4.6, 6.5, and 7.4; solubility at multiple pH points; and pKa with ionic-strength corrections—so that predictions reflect the actual experimental conditions under which compounds will be assayed or formulated.
Many global QSPR models are trained predominantly on drug-like small molecules and may produce unreliable predictions for macrocycles, organometallics, peptides, PROTACs, natural products, or compounds containing uncommon functional groups. BOC Sciences mitigates this by evaluating each prediction against the model's applicability domain, applying local models trained on structurally related compounds when available, and using physics-based methods (DFT, COSMO-RS) that do not depend on training data coverage for novel chemotypes.
Different prediction algorithms often produce divergent estimates for the same compound—one method may predict acceptable solubility while another flags a high risk. BOC Sciences resolves this through consensus scoring that integrates predictions from multiple orthogonal algorithms, weights each method based on its historical performance for the relevant chemical class, and reports both the consensus value and the inter-method spread so that clients understand the degree of prediction uncertainty and can make risk-calibrated decisions.
Computational solubility predictions often assume the compound exists as an ideal supercooled liquid, neglecting crystal lattice energy contributions that can reduce experimental solubility by orders of magnitude. BOC Sciences supplements solution-phase solubility predictions with melting-point-based correction terms derived from group additivity or DSC data, and provides guidance on the expected gap between predicted intrinsic solubility and experimentally measurable crystalline solubility, helping clients interpret computational results in a realistic solid-state context.
BOC Sciences combines molecular modeling with access to experimental chemistry and analytical capabilities. When prediction uncertainty affects a decision, focused measurements can be designed to validate key properties, update local models, or explain outliers instead of leaving computational results disconnected from laboratory evidence.
Prediction panels are selected according to scaffold class, research stage, intended use, available data, and the property trade-offs that matter to the project. This avoids unnecessary endpoints and directs computational effort toward conclusions that can guide compound design or experimental planning.
Standardized structure processing and automated calculation workflows support efficient analysis of compound libraries. Priority rules, output formats, and review checkpoints are agreed before calculation so researchers receive organized rankings and interpretable visualizations suitable for timely selection decisions.
Reports identify model type, endpoint definition, validation evidence, applicability limitations, confidence ranges, and important assumptions. Where models disagree or a structure lies outside supported chemical space, the uncertainty is shown directly rather than hidden behind a single over-precise number.
Client Needs: A medicinal chemistry team was optimizing a series of tertiary amine-containing kinase inhibitors and needed predicted pKa and pH-dependent LogD values for 38 analogues to understand how structural variations around the amine center affected ionization and lipophilicity at physiological pH. Their goal was to maintain sufficient basicity for target engagement while reducing overall LogD to improve metabolic stability.
Challenges: The amine pKa values were sensitive to subtle electronic and steric effects from substituents on the adjacent ring system, and several analogues contained additional ionizable heterocycles that created complex microstate ionization profiles. A simple Hammett-type prediction was insufficient to capture these multi-site effects accurately.
Solution: We applied DFT calculations at the B3LYP/6-311+G(d,p) level with SMD solvation to compute proton affinities and microstate pKa values for all ionizable centers, then combined these with fragment-based LogP predictions to generate pH–LogD profiles from pH 2 to 10. Consensus scoring across three orthogonal LogD algorithms was performed for all 38 compounds, and matched molecular pair analysis identified the substituent modifications that produced the largest shifts in amine basicity and LogD.
Outcome: The client received a ranked series with pKa and LogD7.4 values for each analogue, microstate speciation plots, and specific recommendations for substituent modifications predicted to reduce LogD by 0.5–1.0 log units while preserving target-relevant amine basicity.
Client Needs: A drug discovery group had 25 lead compounds from three distinct chemical series and required pH-dependent aqueous solubility ranking to select the most developable candidates for downstream pharmacokinetic studies. The compounds spanned a wide polarity range and included both neutral and ionizable species.
Challenges: Several compounds exhibited steep pH–solubility curves near the physiological range, meaning small pH differences between prediction conditions and actual assay buffers could produce large discrepancies in solubility estimates. Additionally, two series contained compounds with melting points above 250 °C, suggesting significant crystal lattice contributions to experimental solubility that computational models alone could not capture.
Solution: We generated intrinsic solubility predictions using the general solubility equation (GSE) with melting point correction terms from group additivity, then applied pH-dependent Henderson–Hasselbalch corrections using DFT-calculated pKa values. Full pH–solubility profiles from pH 1 to 8 were generated for all 25 compounds. For the high-melting-point series, we flagged the expected gap between predicted intrinsic solubility and experimental crystalline solubility and recommended experimental solubility measurement for 5 representative compounds to anchor the computational predictions.
Outcome: The client received a ranked solubility table with pH–solubility curves, solubility classification by BCS category, and clear identification of three compounds with solubility below 10 μM at pH 6.5 that were recommended for structural modification or early formulation intervention.
Client Needs: An industrial chemistry team developing volatile organic intermediates needed predicted boiling points, vapor pressures, and Henry's law constants for 60 compounds across three structural families to support process safety assessment and emission control planning. Experimental measurement for the full set was impractical within the project timeline.
Challenges: Two of the structural families contained polyhalogenated and nitro-substituted aromatics with limited representation in standard boiling point and vapor pressure training sets. The third family included flexible aliphatic esters whose vapor pressure was sensitive to conformational population distributions not captured by simple 2D descriptor-based models.
Solution: We applied group additivity methods for boiling point and vapor pressure estimation, supplemented by COSMO-RS calculations for compounds outside the group-contribution parameter space. For the flexible esters, we performed conformational sampling via molecular mechanics followed by DFT single-point energy calculations to generate Boltzmann-weighted 3D descriptors that better captured the conformational ensemble's effect on volatility. Henry's law constants were derived from the predicted vapor pressure–solubility ratio for each compound. In total, 180 property predictions were generated and cross-validated against available literature data for structurally similar reference compounds.
Outcome: The client received predicted boiling points, vapor pressures, and Henry's law constants for all 60 compounds with confidence classifications, enabling risk-based prioritization of experimental verification and supporting process safety documentation for the highest-volatility compounds.
Physicochemical prediction can cover LogP, LogD, pKa, aqueous solubility, molecular weight, polar surface area, hydrogen-bond features, rotatable bonds, permeability, and other property-relevant molecular descriptors. Condition-dependent behavior may also be evaluated at specified pH values, temperatures, ionic strengths, or solvent environments when suitable models and input data are available. The prediction panel is selected according to the chemical structures, research stage, available experimental information, and intended application, so each project can focus on the properties most relevant to subsequent decisions.
An accurate two-dimensional structure or a standard molecular file, such as SMILES, SDF, or MOL, is generally sufficient for initial assessment. Projects involving pH dependence, stereochemistry, tautomerism, salt forms, or specific protonation states should include the corresponding structural and experimental details. Clients may also provide measured property data, relevant analogues, and intended test conditions to support model selection and comparison. Before calculation, structures are reviewed for valence, stereochemistry, duplicates, ionization state, and file consistency to reduce errors caused by incorrect molecular input.
Model selection depends on the target property, number of compounds, structural complexity, available reference data, and required analytical depth. Physics-based or quantum-chemical methods may be suitable for focused questions involving individual compounds, whereas validated QSPR, statistical, or machine-learning models are often more efficient for larger libraries. For novel scaffolds or data-limited chemical series, semi-empirical calculations, similarity analysis, and consensus modeling may be combined. This project-specific approach balances computational speed, chemical interpretability, model coverage, and the level of confidence needed for research decisions.
Prediction reliability is assessed using several factors rather than presenting a calculated value without context. These factors may include the model applicability domain, structural similarity to reference compounds, agreement among independent methods, validation statistics, and consistency with available experimental measurements. Limitations are identified when compounds contain unusual functional groups, fall outside represented chemical space, or may be strongly affected by solid-state behavior. Where supported by the model and dataset, reports may include prediction intervals, uncertainty notes, or confidence categories to guide interpretation and experimental follow-up.
Yes. For compound-library projects, a consistent set of physicochemical properties can be calculated and used to filter, rank, cluster, or compare structures according to project-defined thresholds or multiparameter criteria. The analysis can reveal potential solubility, lipophilicity, ionization, permeability, stability, or reactivity concerns and clarify structure–property trends within an analogue series. These findings can help researchers prioritize compounds for experimental testing, select suitable pH or solvent conditions, and design more focused analytical, formulation, or preformulation studies while reducing unnecessary screening.
BOC Sciences delivered a comprehensive property panel—LogD, pKa, solubility, permeability, and drug-likeness metrics—for our entire lead series in under a week. The consensus scoring approach across multiple prediction methods gave us confidence in the numbers, and the structure–property trend analysis directly informed our next design cycle.
— Dr. Garrison, Principal Scientist, Medicinal Chemistry
What set BOC Sciences apart was the transparency around prediction confidence. Each value came with information about the model used, its applicability domain, and the expected accuracy range. When predictions were uncertain for some of our more unusual chemotypes, they flagged it clearly and recommended where experimental confirmation would add the most value.
— Senior Computational Chemist, Biotechnology Company
The property ranking report made it straightforward to compare over 200 compounds across six different physicochemical parameters simultaneously. The radar charts and property heatmaps were immediately useful in our project team meetings, and the matched pair analysis helped us understand exactly which structural modifications were driving property improvements.
— Dr. Elliott, Lead Discovery Project Leader
Throughout the project, the BOC Sciences team was responsive and scientifically engaged. They asked thoughtful questions about our assay conditions and target property ranges, which led to more relevant predictions. The interpretation call after report delivery was particularly valuable—it turned a data table into a clear set of design recommendations we could act on immediately.
— Head of Discovery Chemistry, Mid-Size Pharma
If you have any questions or encounter issues on this page, please don't hesitate to reach out. Our support team is ready to assist you.
