A novel predictor of ACE2-binding ability among betacoronaviruses

Abstract

We have applied an alignment-free technique to infer functional relatedness among betacoronaviruses. This technique, concurrently being optimized for identifying novel prions, was adapted to gain new insights into coronavirus evolution, specifically in the context of the COVID-19 pandemic caused by novel coronavirus SARS-CoV-2. With COVID-19 already having killed over ∼4.8 million people worldwide as of this writing, novel methods for predicting the capacity for coronaviruses in general to infect human cells are needed. Our proposed method utilizes physicochemical properties of amino acids to develop a fully dynamic waveform representation of proteins that encodes both the amino acid content and the context of amino acids. These waveforms are then subjected to dynamic time warping and distance evaluation to develop a distance metric that is relatively less sensitive to variation in sequence length or primary amino acid composition. Using this method, we show that in contrast to alignment-based maximum likelihood and neighbor-joining phylogenetic analyses, all bat betacoronavirus spike protein receptor binding domains (RBDs) known to bind to the ACE2 receptor are found within a single physicochemical cluster. Further, other RBDs within that cluster are from pangolin coronaviruses, two of which have already been shown to bind to ACE2 while the others are suspected, yet unverified ACE2 binding domains. This finding is important because both SARS-CoV and SARS-CoV-2 use the host ACE2 receptor for cell entry. Surveillance for coronaviruses belonging to this cluster could potentially guide efforts to stifle or curtail potential and/or early zoonotic outbreaks with their associated deaths and financial devastation.

Using machine learning to detect coronaviruses potentially infectious to humans

“…Viruses from non-human hosts according to experimental evidence of binding Binds to human receptor RaTG13 31,35 , LYRa11 34,35 , WIV1 8, 35,38 , WIV16 8, 38 , RsSHC014 34,35 , Rs4231 34 , Rs4084 34,39 , Rs4874 40 , Rs7327 40 , Rs4231 40 , Khosta-1 41 , Khosta-2 41 , Ty-HKU4 24,33 No evidence of binding to human receptor Rc-o319 35 , Bat CoVZXC21 35 , ZC45 34 , HKU5 24,33 , Bat Rf1 35 , Bat Rf4092 35 , Bat BM48-31 34 35 , Bat Rp3 35,44 , bat SL CoV As6526 44 , Bat Rm1 40 , As6526 34 , Yunnan 2011 34 , Shaanxi 2011 34 , Bat 279-2005 34 , Rs4237 34 , Rs4081 34 , Rs4247 34 , Bat HKU3-1 35 , HKU3-8 34 , HKU3-13 34 , GX2013 34 , Longquan-140 34 , YN2013 34 , Rf4092 34 , JL2012 34 , HuB2013 34 , HeB2013 34 , 273-2005 34 The h-BiP score is highly accurate in predicting binding to human receptors for alpha and beta coronaviruses We split the data into training (85%) and testing sets (15%) strati ed by the following groups: hCoV-OC43, hCoV-HKU1, MERS, SARS-CoV-1, SARS-CoV-2, hCoV-NL63, hCoV-229E, other MERS-related, other Sarbecovirus, other Betacoronavirus, porcine epidemic diarrhea virus, other Alphacoronavirus.…”

Using machine learning to detect coronaviruses potentially infectious to humans

“…Binds to human receptor RaTG13 31,35 , LYRa11 34 , 35 , WIV1 8,35,38 , WIV16 8,38 , RsSHC014 34,35 , Rs4231 34 , Rs4084 34,39 , Rs4874 40 , Rs7327 40 , Rs4231 40 , Khosta-1 41 , Khosta-2 41 , Ty-HKU4 24,33 , PCoV_GX 31 , PCoV_GD 31 , A022G 38 , SZ03 38 , B039G 38 , DcCoV UAE-HKU23 42 , Bovine coronavirus isolate Alpaca 32 , camel alphacoronavirus / hCoV229E 29 , MERS Middle East strain 28 , HKU25 33 , YN2018B 39 , Rs9401 39 , Rs3367 39 No evidence of binding to human receptor Rc-o319 35 , Bat CoVZXC21 35 , ZC45 34 , HKU5 24,33 , Bat Rf1 35 , Bat Rf4092 35 , Bat BM48-31 34 35 , Bat Rp3 35,44 , bat SL CoV As6526 44 , Bat Rm1 40 , As6526 34 , Yunnan 2011 34 , Shaanxi 2011 34 , Bat 279-2005 34 , Rs4237 34 , Rs4081 34 , Rs4247 34 , Bat HKU3-1 35 , HKU3-8 34 , HKU3-13 34 , GX2013 34 , Longquan-140 34 The h-BiP score is consistent with sequence identity and expands viral classification We verified that the proposed model, while consistent with sequence alignment results, provides additional information. Percent seq...…”