Design Dissertation Topics

Design Dissertation Topics

Info: Design Dissertation Topics
Published: 15th July 2025 in Design Dissertation Topics

Share this:

Dissertation Topic 1:

Reconceptualizing Structure-Based Drug Design: A Practical Approach to Evaluating SBDD Models

Background Context

Structure-based drug design (SBDD) has provided an exciting and informative way to speed up drug discovery efforts by using computational models to predict molecular interactions. While current metrics used to evaluate SBDD mechanistically rely on Vina docking scores and tend to overestimate the utility of molecules designed, while concentrating on theoretical relevance rather than biological relevance (Guan et al., 2024), it must be emphasised that these models do well with optimising docking scores, but poorly in the important areas of molecular similarity compared to known active compounds, and with virtual screening (Huang et al., 2024). These gaps are important to focus on, which is what this work intends to establish, and it is aimed at creating a new generation of more rigorous evaluation metrics to move beyond strictly Vina-style measurements of how well SBDD models will meet the requirements of drug discovery next steps.

Research Questions

1. How can new evaluation metrics be developed to assess practical applicability in drug discovery, beyond Vina docking scores?
2. In what ways do alternative molecular descriptors, such as delta score and similarity to known actives, improve SBDD model predictive accuracy?
3. How do SBDD models strike a balance between producing high docking scores and producing molecules that are metabolically diverse and biologically relevant?

Potential Implications

  • Better Drug Discovery Models: Improved evaluation metrics could generate an SBDD model that is more reflective of the real world, leading to drug discovery models that are more reliable.
  • Less Overfitting: Eliminating the use of docking scores may mitigate the problem of overfitting, leading to improved generalisability of the models.
  • Faster Validation: More accurate models may reduce the time needed to proceed to experimental testing, thereby speeding up the timeline of drug development.
  • Suggested Reading

  • Gao, B., Tan, H., Huang, Y., Ren, M., Huang, X., Ma, W. Y., … & Lan, Y. (2025, January). Reframing Structure-Based Drug Design Model Evaluation via Metrics Correlated to Practical Needs. In The Thirteenth International Conference on Learning Representations.
  • Gangwal, A., & Lavecchia, A. (2025). Artificial intelligence in natural product drug discovery: current applications and future perspectives. Journal of medicinal chemistry, 68(4), 3948-3969.
  • Would you like to extend the threshold for drug discovery by advancing the practical usefulness of SBDD models?
    Contact us to explore how your research can help improve computational drug design.

    Dissertation Topic 2:

    Evaluating the Impact of Molecular Diversity in Structure-Based Drug Design: Beyond Vina Scores

    Background Context

    Although Structure-Based Drug Design (SBDD) models have been improved to yield high-scoring docked molecules, the evidence for structural diversity of the generated molecular structures lacks substance. Many SBDD models are still focused blindly on a high-scoring docked molecule, which ultimately leads to the plus of molecules having low structural diversity, which leads to low molecular opportunities in drug discovery (Guan et al., 2024). Diverse molecular libraries need to be built to enhance the understanding of the greater therapeutic value of SBDD-generated compounds. The present study will attempt to evaluate the value of molecular diversity in SBDD models and develop different metrics which can incorporate diversity and performance into docking models.

    Research Questions

    1. How can the molecular diversity of SBDD models be quantitatively determined, and what is the value of this molecular diversity in the context of drug discovery?
    2. What is the value of modelling accuracy associated with adding molecular diversity models to the predictions for therapeutic efficacy versus only using the docking scores?
    3. How can SBDD models be improved to take account of both molecular diversity and docking score?

    Potential Implications

  • Better Libraries: Looking for molecular diversity will result in stronger compound libraries for drug screening.
  • Better Affinity Predictions: With diversity metrics built into models, we could generate molecules that may achieve better targeted efficacy.
  • Better Thunderbird Discovery: Having more molecular variety could discover new drug candidates for diseases with unmet needs.
  • Suggested Reading

  • Siddiqui, B., Yadav, C. S., Akil, M., Faiyyaz, M., Khan, A. R., Ahmad, N., … & Azad, I. (2025). Artificial intelligence in computer-aided drug design (cadd) tools for the finding of potent biologically active small molecules: Traditional to modern approach. Combinatorial Chemistry & High Throughput Screening.
  • An, Q., Huang, L., Wang, C., Wang, D., & Tu, Y. (2025). New strategies to enhance the efficiency and precision of drug discovery. Frontiers in Pharmacology, 16, 1550158.
  • Would you like to incorporate molecular diversity into your SBDD models to help advance drug discovery and development?
    Get in touch with us to discuss how your research can influence the advancement of more accurate methods for drug screening.

    Dissertation Topic 3:

    Optimizing Virtual Screening in SBDD: A New Metric for Simulating Drug-Ligand Interactions

    Background Context

    Virtual screening is an integral aspect of SBDD. It predicts how a target protein and drug candidates might interact. Still too often, the SBDD models focused too much on docking score as the optimal metric to estimate efficacy and thus missed a large array of drug-ligand interactions that may be critical in design (Huang et al. 2024). Newer metrics that could simulate better drug-ligand adhesion dynamics, thus refining virtual screening model performance, may help to identify the molecules with the highest potential early in the discovery process. The goal of this dissertation topic is to develop a better metric for virtual screening that would induce the drug-ligand interaction dynamics to be more realistic.

    Research Questions

    1. What are the main factors affecting drug-ligand interactions not considered in docking scores?
    2. How can advanced metrics for simulating drug-ligand interactions improve the process of virtual screening in SBDD?
    3. What is the effect of this new virtual screening method on the overall performance of the SBDD models to find active drug candidates?

    Potential Implications

  • Improved Virtual Screening: improved simulations of drug-ligand interactions could improve accuracy in drug discovery, during the earlier stages of drug development.
  • Improved Candidate Identification: faster identification of biologically active compounds.
  • Improved SBDD Model Predictions: improved predictive capacity of SBDD models could improve drug discovery.
  • Suggested Reading

  • Inglot, T. W. (2025). Exploring pitolisant binding to human serum albumin: Insights from multi-spectroscopic, molecular docking and molecular dynamics studies. Journal of Molecular Structure, 142698.
  • Shrivastav, P., Prajapati, B., Chandarana, C., & Prajapati, P. (2025). In Silico Modeling of Drug–Receptor Interactions for Rational Drug Design in Neuropharmacology. Computational Neuropharmacology: Fundamentals and Clinical Aspects, 87-126.
  • Would you like to improve virtual screening methods in drug discovery?
    Contact us to talk about how your research can change the simulation of drug-ligand interactions and overall drug development timelines.

    Dissertation Topic 4:

    Incorporating Biological Relevance into Structure-Based Drug Design: A Holistic Approach to Model Validation

    Background Context

    SBDD models routinely employ docking scores as theoretical metrics without verifying the biological relevance of the resulting molecules (Guan et al., 2024). For any drug discovery model to be valid, its goal must be not only to estimate the binding affinity of a molecule but also to estimate what it employs molecularly when interacting with a biological target as a deviant drug-like molecule. This research seeks to assess the biological relevance of evaluating SBDD models and a framework that assesses the molecular and biological properties of the produced compounds.

    Research Questions

    1. How can biological relevance be used when evaluating SBDD models?
    2. What biological metrics (e.g., target engagement, toxicity, metabolic stability) need to be considered to validate compounds generated by SBDD?
    3. How can biological relevance improve the overall success rate of drug discovery in SBDD models?

    Potential Impact

  • Model Validation Improvement: Achievement of understanding biological metrics and therefore developing better models, which are sustainable with real-world drug discovery pathways.
  • Therapeutic Success: Though increasingly biologically relevant SBDD models and then drug discovery candidates could improve efficacy and safety.
  • Efficiency in the Drug Discovery Process: This could also lead to increased rates of success and fewer candidates failing in clinical trials because problems would have been detected earlier in the drug development cycle.
  • Suggested Reading

  • Mharazanye, D. K., & Muskaan, S. K. Modern Strategies for Cns Drug Discovery: Integrating Cadd And Deep Learning For Therapeutic Advances. REDVET-Revista electrónica de Veterinaria, 26(1), 2025.
  • Mharazanye, D. K., & Muskaan, S. K. Modern Strategies for Cns Drug Discovery: Integrating Cadd And Deep Learning For Therapeutic Advances. REDVET-Revista electrónica de Veterinaria, 26(1), 2025.
  • Do you want to improve drug discovery through biologically relevant SBDD models?
    Reach out to us to discover how we can add biological metrics to your designs, improve your research framework and more accurately fit your models to the objectives of your study.

    Dissertation Topic 4:

    Incorporating Biological Relevance into Structure-Based Drug Design: A Holistic Approach to Model Validation

    Background context

    One of the larger issues with contemporary SBDD models is the overfitting of molecules to docking scores, ultimately leading to compounds that may not be viable drug candidates (Huang et al., 2024). Many models may become too reliant on docking score optimisation that leads to the development of molecules that ultimately fail to become viable drug candidates in later stages of drug development due to, most importantly, poor drug-like properties (i.e., low bioavailability, metabolically unstable). This dissertation is to develop a set of robust evaluation metrics for incorporating key drug-like properties aside from docking scores to mitigate the chances of overfitting.

    Research Questions

    1. What is the key drug-like properties (e.g., ADMET, solubility) that incorporation into current SBDD models has not effectively done?
    2. How do new evaluation metrics improve overfitting and provide molecules that have drug-like properties?
    3. How does concentrating on drug-like properties affect the success of SBDD models in finding viable drug candidates?

    Potential Implications

  • Lower Potential for Overfitting: If the focus were switched away from docking scores, this could reduce overfitting to these scores and produce more reliable and more varied drug candidates.
  • Better Drug-Like Property Prediction: If drug-like properties were incorporated into the evaluation framework, this would increase the chances of successful clinical translation.
  • Better SBDD Models: This method could result in better models creating compounds with increased potential as real-world drugs.
  • Suggested Readings

  • Hu, X., Liu, G., Chen, C., Zhao, Y., Zhang, H., & Liu, X. (2025). TransDiffSBDD: Causality-Aware Multi-Modal Structure-Based Drug Design. arXiv preprint arXiv:2503.20913.
  • Bhattacharjee, T., & Bhatia, R. (2025). Advancements in Structure-based Drug Design Using Geometric Deep Learning. Current Medicinal Chemistry.
  • Do you want to enhance SBDD models incorporating drug-like properties?
    Get in touch with us to find out how your research could help reduce overfitting through our approach and identify potential drug candidate

    Share this:

    Cite this work

    Study Resources

    Free resources to assist you with your university studies!

    This will close in 0 seconds