HIV Drug Resistance prediction using in-silico methods!

Saptaparna Dasgupta, Bennett University

A genotype-based approach for predicting HIV medication resistance would considerably aid in the selection of optimal antiviral drug combinations. Lack of information on the 3D structure of the protein and the processing of partial sequencing data were two key drawbacks of prior models. Homology modeling and molecular field mapping might be used to get 3D structural information on viral proteins. As a result, drug resistance prediction models, using these data analysis approaches on the HIV-1 protease inhibitor dataset were created.

Prediction models: Why are they important?

The influence of HIV drug-resistant viruses has a great impact on HIV prognosis. Using their genetics to predict medication resistance would allow for more effective therapy. Several prediction models have been developed, and databases, such as the Stanford HIV Drug Resistance Database that gathers protein information and analyses the resistance of drug-resistant viruses, are expanding. These models may be classified or regressed using a variety of machine learning techniques, such as support vector machine, deep learning, etc. A protein’s main sequence is transformed into numerical descriptors using one-hot encoding or similar approaches and utilized as predictive variables in the models. Through the use of deep learning, Margaret et al. created a classification model with high accuracy of about 0.9.

The new insights of homology modeling:

Homology modeling, structural alignment, and molecular field mapping were used to encode the structural information of HIV-1 variations as 3D molecular field parameters. Based on comparative molecular field analysis (CoMFA), molecular field mapping is a technique. To compute the interaction between the probe atoms and the molecule, CoMFA embeds the molecule of interest in a grid lattice. Molecular field parameters computed are utilized as characteristics linked with each molecule’s physical structure. Normally, CoMFA is utilized for QSAR analysis of low-molecular-weight medicines, but in this work, it was used in the other way to the homologous drug target (i.e., viral proteins).

The study:

Stanford University’s HIV Drug Resistance Database provided data on HIV-1 drug resistance as well as the main structure of HIV-1 protease variants. The fold change (FC) increase in the IC50 of HIV-1 protease inhibitors relative to wildtype HIV was a measure of resistance. The FC threshold was set at 3.5, and viruses with a higher score were classed as resistant. Incomplete sequencing samples were used in the data collection, resulting in numerous possibilities for some amino acid locations. On the basis of their probabilities, the uncertainty of all combinations was evaluated. A version of Modeler 9.20 was used for the homology modeling. A template was created using HIV-1 protease crystal structures found in the PDB library. For alignment of the main sequence with the templates, mufft v7.427 was used. Followed by which the prediction models were constructed, and the sample-based selection was made and the hyperparameters were optimized.

Results obtained:

Since not all viral types were examined, the sample sizes differed across the medications. Homology modeling was used to estimate the 3D structure of each HIV-1 protease variant. Each medication was predicted to have a large number of samples. The results were about twice as informative as those obtained from sequencing samples that were completed. A training dataset and an external test dataset for each medication were created. To develop the prediction models, PLS, LGBM, RF, and SVR were chosen. Compared to Geno2pheno, the current regression model had a higher predictive R2 (0.698). Drug resistance is acquired through viral mutation, and a 3D-based study provides a reasonable representation of the structure-activity connection. Amino acids known to cause drug resistance are located around the contour maps. 

Future of the technique and observed limitations:

Based on the 3D structure of HIV protease variations, a predictive model for HIV medication resistance with good prediction accuracy was successfully built. Other anti-HIV drugs such as reverse transcriptase inhibitors and integrase inhibitors can also be predicted using the suggested technique. However, it must also be noted that this technique would be difficult to implement in an environment that is restricted in computing resources since reverse transcriptase and integrase are much larger than protease.

Also read: Impact of plastic pollution on marine turtles

Reference:

1. Ota, R., So, K., Tsuda, M., Higuchi, Y., & Yamashita, F. (2021). Prediction of HIV drug resistance based on the 3D protein structure: Proposal of molecular field mapping. PLOS ONE, 16(8), e0255693. https://doi.org/10.1371/journal.pone.0255693

• The Corrosion Prediction from the Corrosion Product Performance
• Nitrogen Resilience in Waterlogged Soybean plants
• Cell Senescence in Type II Diabetes: Therapeutic Potential
• Transgene-Free Canker-Resistant Citrus sinensis with Cas12/RNP
• AI Literacy in Early Childhood Education: Challenges and Opportunities

Author info:

Saptaparna Dasgupta, currently a B. Tech 3^rd year student, pursuing Biotechnology, is a diligent student and determined in terms of her career goals. Being a budding biotechnologist, she is open to all research fields of her course and passionate about knowledge. She is focused and constantly tries to improve her writing skills, also a project enthusiast and is fond of gaining the hands-on experience in laboratories. She believes that all hard works and efforts pays off eventually and follows this as the motto of her life.