“…As summarized in Table 4 (Tab. 4) (References in Table 4: Nagy et al, 1994[ 58 ]; Recanatini, 1996[ 75 ]; Oprea and García, 1996[ 67 ]; Sulea et al, 1997[ 92 ]; Recanatini and Cavalli, 1998[ 76 ]; Cavalli et al, 2000[ 18 ]; Beger et al, 2001[ 10 ]; Gironés and Carbó-Dorca, 2002[ 34 ]; Beger and Wilkes, 2002[ 12 ]; Polanski and Gieleciak, 2003[ 71 ]; Leonetti et al, 2004[ 50 ]; Beger et al, 2004[ 11 ]; Cavalli et al, 2005[ 17 ]; Bak and Polanski, 2007[ 8 ]; Castellano et al, 2008[ 16 ]; Nagar et al, 2008[ 55 ]; Mittal et al, 2009[ 53 ]; Gueto et al, 2009[ 36 ]; Dai et al, 2010[ 24 ]; Roy and Roy, 2010[ 78 ]; Roy and Roy, 2010[ 77 ]; Nagar and Saha, 2010[ 57 ]; Nagar and Saha, 2010[ 56 ]; Narayana et al, 2012[ 65 ]; Nantasenamat et al, 2013[ 64 ]; Nantasenamat et al, 2013[ 61 ]; Kishore et al, 2013[ 44 ]; Worachartcheewan et al, 2014[ 103 ]; Worachartcheewan et al, 2014[ 101 ]; Nantasenamat et al, 2014[ 63 ]; Dai et al, 2014[ 25 ]; Awasthi et al, 2015[ 7 ]; Xie et al, 2015[ 105 ]; Shoombuatong et al, 2015[ 85 ]; Xie et al, 2014[ 102 ]; Kumar et al, 2016[ 48 ]; Ghodsi and Hemmateenejad, 2016[ 32 ]; Song et al, 2016[ 91 ]; Prachayasittikul et al, 2017[ 72 ]; Adhikari et al, 2017[ 1 ]; Lee and Barron, 2018[ 49 ]; Pingaew et al, 2018[ 70 ]; Barigye et al, 2018[ 9 ]), it can be observed that prior to 2010, MLR and PLS models, also known as white-box approaches, were the most popular and yet simple learning algorithms used for QSAR modeling of AIs.…”