Microbiological Seminar, Brno, 2021
Artificial Intelligence
in Microbiology
by St anislav Maz urenko, PhD
maz urenko@mail.muni.cz
Outline
 Motivation
 Introduction to AI and ML
 Recent applications in Microbiology
2/22
Motivation
Motivation: sequences and chemicals
 Large volumes of digital data
 Affordable computing power and storage
 Complex study objects
4/22
0.1
1
10
100
1000
10000
1993 1998 2004 2009 2014 2020
Millions
The total number of sequences
GenBank
WGS
5/22
Source: Scheler et al. "Recent developments of microfluidics as a tool for biotechnology and microbiology." Current opinion in biotechnology 2019 Khater
et al. "Picoliter agar droplet breakup in microfluidics meets microbiology application: numerical and experimental approaches." Lab on a Chip2020.
Motivation: big experimental data
6/22Source: Scher et al. "In-situ fiducial markers for 3D correlative cryo-fluorescence and FIB-SEM imaging." iScience (2021): 102714.
Motivation: cell imaging
Introduction to AI and ML
 Recommendation engines
 Image & speech recognition
 Anomaly detection
 Natural language processing
 Data mining…
Introduction to AI and ML
8/22
Source: towardsdatascience.com
Introduction to AI and ML
9/22
Introduction to ML
 Historically, people tried to find rules themselves, e.g. detection of particular
shapes or color contrasts;
 Often such manual rules are too simplistic to give good results;
 Machine Learning gives the means to generates those rules automatically!
Faces Not faces
Non-vaccine candidates: Vaccine candidates:
10/22
F( ) = + 1 F( ) = - 1
Basics of ML: data representation
MKKLGRAATNKAAKEVLDYCGEAKG… Feature vector: (5, 1, 1, -5.67, 0.69, …)
One-hot encoding:
K K L G R A A T …
A 0 0 0 0 0 1 1 0 …
K 1 1 0 0 0 0 0 0 …
L 0 0 1 0 0 0 0 0 …
G 0 0 0 1 0 0 0 0 …
R 0 0 0 0 1 0 0 0 …
T 0 0 0 0 0 0 0 1 …
…
11/22Source: Goodswen et al. "Machine learning and applications in microbiology." FEMS Microbiology Reviews (2021).
Examples:
• AA frequency
• AA sequence
• Conservation scores
• Structural elements
• …
12/22
F( ) = -1
Basics of ML: training
0.3
1.4
1
-1
F( ) = +1
13/22
Multiple
parameters
Basics of ML: training
1.2
0.5
10
+1
Basics of ML: validation
Data
Train an ML predictor
Test data Training data
Evaluate the predictor
14/22
 The goal of ML is to identify generalizable patterns in your training data.
 These patterns must be valid for future data!
 Therefore, the core of ML protocol is to evaluate the predictor on the test
data, hidden from the predictor:
Artificial Neural Networks
15/22Link: https://vimeo.com/154085950
Recent applications
Source: Peiffer-Smadja et al. "Machine learning in the clinical microbiology laboratory: has the time come for routine practice?"
Clinical Microbiology and Infection 2020
Overview
17/22
Bacterial colony morphology
Source: Huang, Lei, and Tong Wu. "Novel neural network application for bacterial colony classification." Theoretical Biology and
Medical Modelling 15.1 (2018): 1-16. 18/22
Interclass variations Intraclass variations (Streptococcus agalactiae)
A convolutional neural network was able to discriminate between 18 classes of bacterial colonies.
Source: Ho Chi-Sing et al. "Rapid identification of pathogenic bacteria using Raman spectroscopy and deep learning."
Nature communications 10.1 (2019): 1-8.
Identification of pathogens
19/22
Source: Khaledi Ariane et al. "Predicting antimicrobial resistance in Pseudomonas aeruginosa with machine learning‐enabled
molecular diagnostics." EMBO molecular medicine 12.3 (2020): e10264.
Antimicrobial resistance
20/22
Source: C4X Discovery, Fernández-Torras et al. "Connecting chemistry and biology through molecular descriptors."
Current Opinion in Chemical Biology 66 (2022): 102090.
Clinical outcomes
21/22
Summary
 Machine Learning method is a powerful data-driven alternative to
traditional modelling;
 One turns data into numbers (features) and trains a generic
algorithm to discriminate between labels in the feature space;
 It is essential to have a separate test set for evaluation of the
resulting predictor;
 In Microbiology, a wide range of tasks is already solved by
Machine Learning.
22/22
Bi9680En: AI in Biology, Chemistry, and Bioengineering
 Období: podzim
 Rozsah: přednáška 2 hodiny/týden
 Vyučující: Dr. Stanislav Mazurenko
 Osnova:
 modern bio-challenges: drug design, DNA interpretation, protein engineering
 types of AI algorithms and workflow for designing predictors
 clustering algorithms, random forests, artificial neural networks
 features, databases, and predictors used in applications