Studies estimate the effect of amino acid changes on SARS-CoV-2 infection

Costas D at Pennsylvania State University. A new study led by Maranas predicts amino acid changes to the receptor-binding domain of Severe Acute Respiratory Syndrome coronavirus 2 (SARS-CoV-2). Spike protein It adversely affects binding affinity and subsequent infection of human cells.

Their results are a new two-step procedure called Neural Network Molecular Mechanics / Poisson-Boltzmann Surface Area (NN_MM-GBSA) that calculates the binding energy from a receptor binding domain variant to the human angiotensin converting enzyme 2 (ACE2) receptor. Derived from. The second step is to build a neural network from the findings to predict the binding affinity. The team achieved an accuracy rate of 82.2% by classifying amino acid substitutions as useful or unhelpful in the binding affinity of the variant.

The researcher wrote:

“Therefore, our method sets up a framework for effectively screening for binding affinity changes with unknown single and multiple amino acid changes. This is the current and future SARS-CoV-. 2 It is an invaluable tool for host adaptation of mutants and for predicting zoonotic outbreaks. “

The study “Calculation Prediction of the Effect of Amino Acid Changes on Binding Affinities between SARS-CoV-2 Peplomers and Human ACE2 Receptors” bioRxiv* Servers and articles have been peer reviewed.

(A) Crystal structure of the complex formed between the RBD and the hACE2 complex. The ACE2 protein is shown in blue as a surface representation and the RBD is shown in magenta. (B) Residues of the RBD mutant in direct contact with hACE2 are depicted as cyan spheres. Residues that are not in direct contact are orange. (C) Histogram showing the ratio of KDs and apps for all 108 RBD variants in the dataset. The black histogram bar shows the number of variants in the training set that increase the binding affinity compared to WT (KD, app ratio> 1.0), and the gray bar shows the number of variants that decrease the binding affinity. (KD, app ratio <1.0).

research result

The team used a set of 108 variants to assess the binding energy and affinity of the receptor binding domain of the coronavirus variant. The predictions were based on the MM-GBSA binding energy, which is in partial agreement with the data. We then used binding energy to train a neural network regression model using the degraded MM-GBSA energy term and the dissociation constant ratio between the coronavirus mutant and the original wild-type strain.

“This study reported an apparent dissociation constant K._{D, app} Percentage of all variants in which a single amino acid can change at all RBD positions. AK_{D, app} Ratio (ie K_{D, app, WT}/ K_{D, app, variant}) Variants greater than 1 mean stronger bonds compared to WT, values ​​less than 1 mean weaker bonds. “

NN_MM-Schematic of workflow for building a GBSA model. (A) MD simulations were performed for each single point amino acid substitution variant in the explicit solvent to calculate the decomposition components of the binding energy, followed by MM-GBSA analysis. (B) The MM-GBSA binding energy component was supplied as an input to the neural network using the Kd, app ratio as the regression target. The model is trained using 5 cycles of a 5-fold cross-validation procedure.

Their model had a Pearson correlation coefficient of 0.69 between the predicted and values, with an 82% accurate prediction of the effects of amino acid substitutions and their binding, showing robust accuracy in the results. No trend lines were observed in the training dataset as they continued to add more variants.

The weakest binding affinity appeared to occur when a single amino acid substitution was far from the receptor binding domain of the peplomer.

Amino acid changes in the E484K, N501Y, and K417N mutations from the South African B.1.351 mutant, and the E484K, N501Y, and K417T mutations in the Brazilian P1 mutant increased binding affinity for the human ACE2 receptor. Researchers note that K417N and K417T are known to reduce binding affinity. However, given that the mutants become more infectious, researchers suggest that E484K and N501Y dominate the virus and have a greater impact on binding affinity.

The team then examined variants with double mutations that were suspected to be more infectious. They evaluated the binding affinity of the two mutants with either V503W and E406F or V503W and Y505W. The mutant had a large hydrophobic amino acid that contributed to a higher binding affinity with the human ACE2 receptor. Hydrophobic interactions are primarily associated with solid anchors for binding.

NN_MM-GBSA limits

Researchers have pointed out a serious drawback of the two-step procedure. It is very time consuming and relies on theoretical inferences about binding energy and subsequent affinity for human ACE2 receptors. This limitation can be circumvented by making new predictions using existing energy terms from a balanced training set of 108 variants, but the neural network model will need to be retrained.

Future work using a two-step procedure will allow us to focus on cases of animal coronavirus. Previous studies have shown evidence that SARS-CoV-2 infects cats, dogs, and ferrets, but it is unclear how mutants affect these species. “Assuming that training for NN_MM-GBSA with hACE2 data is robust, it can, in principle, be used to positively assess the relative affinity of RBD for circulating variants for various animal ACE2.”

Important Notices

* bioRxiv Publish preliminary scientific reports that should not be considered definitive as they are not peer-reviewed, guide clinical practice / health-related behaviors, and should not be treated as established information.