ACAI

Aquatic Chemistry with Artificial Intelligence

HomePage

Traditional Aquatic Chemistry

Intelligent Aquatic Chemistry

Introduction

Antibiotic
Resistance
Genes

Coronavirus

Organic
Contaminats

Disinfection
by-products

■ Introduction

This is a free online system for automatic prediction of oxidation, adsorption, and other processes related to aquatic chemistry based on machine learning models, which were trained with data extracted from published articles. Users are welcome to input necessary variables to start the prediction for the parameters of interest, thus providing them with a decision-making reference for their experimental design and parameter selection.

■ Introduction

In this page, machine learning models were developed to predict the degradation and deactivation rates of ARGs by disinfection with chlorine, ultraviolet light, ozone, and hydrogen peroxide, based on data collected from the published papers. Users can predict the degradation rates of ARGs determined with qPCR technology and the deactivation rates of ARGs determined with transformation assay in the above disinfection processes following the prompts. More detailed information about the development of machine learning models is available in the following study.

Ruixing Huang, Chengxue Ma, Xiaoliu Huangfu. A framework predicting removal efficacy of antibiotic resistance genes during disinfection processes with machine learning.

■ ARG degradation prediction

● Oxidants:

● Inputs:

extracellular or intracellular? 1 for extracellular and 0 for intracellular:

initial concentration of plasmid/chromosome (copies/mL) or bacteria (CFU/mL):

length of amplicon (bp):

exposure of FAC (Cl2) (M·s):

temperature(°C):

pH:

single stranded gene sequence of the targeted ARG:

the ARG degradation rate is predicted to be Log(C/C0) =

■ ARG deactivation prediction

ARG deactivation prediction is not categorized by the type of disinfectant, the following model applies to the four disinfection processes with chlorine, ultraviolet light, ozone, or hydrogen peroxide.

● Inputs:

if FAC disinfection? 1 for FAC and 0 for not:

if UV disinfection? 1 for UV and 0 for not:

if O3 disinfection? 1 for O3 and 0 for not:

if H2O2 disinfection? 1 for H2O2 and 0 for not:

repair-deficient mutation of the recipient cell, 1 for recA-:

repair-deficient mutation of the recipient cell, 1 for uvrA-:

repair-deficient mutation of the recipient cell, 1 for endA-:

concentration of the recipient cells (cells/mL):

extracellular or intracellular? 1 for extracellular and 0 for intracellular:

length of amplicon (bp):

degradation rate of the tested ARGs (Log(c/c0)):

initial concentration of plasmid/chromosome (copies/mL) or bacteria (CFU/mL):

pH:

concentration of tert-butanol (mM):

single stranded gene sequence of the targeted ARG:

the ARG deactivation rate is predicted to be Log(C/C0) =

■ ARG degradation-deactivation tandem prediction

● Oxidants:

● Inputs:

extracellular or intracellular? 1 for extracellular and 0 for intracellular:

initial concentration of plasmid/chromosome (copies/mL) or bacteria (CFU/mL):

length of amplicon (bp):

exposure of FAC (Cl2) (M·s):

temperature (°C):

pH:

repair-deficient mutation of the recipient cell, 1 for recA-:

repair-deficient mutation of the recipient cell, 1 for uvrA-:

repair-deficient mutation of the recipient cell, 1 for endA-:

concentration of the recipient cells (cells/mL):

single stranded gene sequence of the targeted ARG:

the ARG degradation rate is predicted to be Log(C/C0) =

the ARG deactivation rate is predicted to be Log(C/C0) =

■ Introduction

In this page, machine learning models were developed to predict the environmental stability and UV inactivation of coronavirus, in particular, SARS-Cov-2. For the environmental stability of coronavirus, the decay of coronavirus titers in solutions and on surfaces over time at -20 to 100 °C can be predicted. For UV inactivation of coronavirus, the decay of coronavirus titers in solutions and on surfaces with UV dose at wavelengths of 222-413 nm can be predicted. More detailed information about the development of machine learning models is available in the following study. A framework predicting removal efficacy of antibiotic resistance genes during disinfection processes with machine learning.

Ruixing Huang, Chengxue Ma, Xiaoliu Huangfu. Preparing for the Next Pandemic: Predicting the Environmental Stability of Coronavirus with Machine Learning.

Ruixing Huang, Chengxue Ma, Xiaoliu Huangfu, Jun Ma. Preparing for the Next Pandemic: Predicting UV Inactivation of Coronaviruses with Machine Learning. Environ. Sci. Technol. 2023, 57, 13767−13777

■ Coronavirus environmental stability prediction

● Scenario:

● Inputs:

the concentration of FBS in the culture medium:

number of days post virus inoculation into the medium before its introduction into solution:

temperature of the background solution (°C):

pH of the background solution:

the length of time the virus is present in solution (h):

the logarithm of viral initial titers (Log TCID50/mL):

gene sequence length of the virus (bp):

single stranded gene sequence of the targeted coronavirus:

the residual virus titer is predicted to be (Log TCID50/mL):

■ Coronavirus UV inactivation prediction

● Scenario:

● Inputs:

wavelength of UV (nm):

exposure of UV (mJ/cm2):

the petri dish is glass or plastic? 1 for glass and 0 for plastic:

the culture medium is DMEM or not? 1 for yes and 0 for not:

the concentration of FBS in the culture medium:

number of days post virus inoculation into the medium before its introduction into solution:

the logarithm of viral initial titers (Log TCID50/mL):

gene sequence length of the virus (bp):

single stranded gene sequence of the targeted coronavirus:

the residual virus titer is predicted to be (Log TCID50/mL):

■ Introduction

This Python simulator for second rate constant was built based on MF-LightGBM and the MD-XGBoost models. On this page, descriptors (MF and MD) and radical types (chlorine radical, dichloro radical, bromine radical, and dibromo radical) are provided as input feature options. The reactivity of halogen radicals with aqueous organic pollutants can be rapidly predicted by entering the SMILES, pH and temperature.

■ Rate constant prediction

● Radicals type:

● Feature type:

● Inputs:

SMILES

pH

T

The rate constant is predicted to be LogK=

■ Introduction

This Python simulator enables the prediction of DBP formation in drinking water treatment, including total DBPs (THM4, DHAA3, THAA4, DBAN3, HK2) and individual DBPs (TCM, BDCM, CDBM, DCAA, TCAA, DBAA, DCAN, BCAN, 1,1-DCP, and 1,1,1-TCP). Users need to provide the DBP type, water quality parameters (DOC, UV254, SUVA, and Br_CONC), operational conditions (Method, Dose, pH, Temp, and Time), or/and the SMILES string (for individual DBPs) to get the predicted DBP level. More detailed information about the development of machine learning models is available in the following study.

Youheng Liang et.al, Machine learning-guided prediction of chlorinated/chloraminated disinfection by-product formation in drinking water treatment.

■ DBPs_CONC prediction

● DBPs type:

● Inputs:

The predicted Log (DBPs_CONC) (μg/L) =

The predicted DBPs_CONC (μg/L) =