/ 85

A diversity-aware computational framework for

systems biologyPhD Candidate: Roberta Bardini

Advisor: Prof. Alfredo Benso

PhD Program: Control and Computer EngineeringXXXI cycle

Turin, September 18th, 2019

/ 85

Activities overview

Scientific research and projects

Soft skills and teaching activities

Diversity-aware computational framework for systems biologyHeartVdA national project

ScRNA-seq analysis

Pending patent for improving cell culture

MITOR project for modeling diffusion

I4C courseErickson school

Training and coaching for SEITraining at TUM

Tutoring Tech4Disability

2015 2019

2

/ 85

Activities overview

Diversity-aware computational framework for systems biology

HeartVdA national project

ScRNA-seq analysis

Pending patent for improving cell culture

MITOR project for modeling diffusion

I4C courseErickson school

Training and coaching for SEITraining at TUM

Tutoring Tech4Disability

2015 2019

Scientific research and projects

Soft skills and teaching activities3

/ 85

Thesis statement

Different scientific domains contribute to systems biology.

A unified computational language for modeling biological systems could enable interdisciplinary collaboration.

The proposed computational framework includes a modeling

approach for complex biological systems, and a model description

language to make it usable for the diverse systems biology community.

4

/ 85

Outline

1. Biological complexity

2. Systems biology as a method

3. Systems biology as a domain

4. Managing knowledge

5. A diversity-aware computational framework for systems biology

6. Future perspectives

5

/ 85

1. Biological complexity

6

/ 85

Biological systems are complex and adaptive

Biologicalentities

Biologicalbehavior

Biologicalinteractions

Complex biologicalpatterns emerge

from localinteractions

7

/ 85

Biological systems are multi-level

A biological entitycan work as "the

changing localenvironment" to

other ones8

/ 85

Biological local interactionsPotential

interactionsActualinteractions

Spatiality

Spatiality mediatesfunctional

interactions.9

/ 85

Ontogenetic processes

© Ralph A. Clevenger/Getty

10

/ 85

Basic developmental mechanisms

Adapted from Salazar et al, 2003

Local interactions occur at different

system levels11

/ 85

Morphogens gradients...

Morphogengradients encodesignals through

distance12

/ 85

...over complex shapes

Salazar et al, 2003

Multiple gradientsin 3D space

generate complexarchitectures

13

/ 85

Process organization

Salazar et al, 2003

Local interactions at different levels

take turns in temporal phases

14

/ 85

Morphogenetic patterns

15

/ 85

A biological system taking shape

Colette et al, 2015

Architectural

complexityemerges alongdevelopment

16

/ 85

Cell fates and differentiation

Takahashi and Yamanaka, 2015

Phenotype

complexityemerges alongdevelopment

17

/ 85

A dynamic landscape of regulations

Noble, 2015 (modified from Waddington, 1957)

Genetic and epigeneticregulations draw a

flexible hierarchy of regulations over morphogenesis.

18

/ 85

2. Systems biology as a methodBotanical Notebook of Linnaeus, ca 1751, World Digital Library

19

/ 85

Multiple organization levels

Systems biologyorganizes

reductionisticexplanations in

schemes of relationships...

20

/ 85

The systems biology circular pipeline

...embeddingquantitative data

into functionalrepresentations to

generate new hypothese to test

in the lab.21

/ 85

The systems biology circular pipeline

This work focuseson computational

approaches for systems biology.

22

/ 85

3. Systems biology as a domain

IBM Research, 2019

23

/ 85

Different types of information

Yugi et al, 2016

Consideringmultiple system

levels impliescomprising

information from different biological

branches and omics...

24

/ 85

...cultures (evenwithin the same

group of experts)...

Different scientific cultures

25

/ 85

...and languages.

Different ways to express knowledge

26

/ 85

4. Managing knowledge

27

/ 85

Models to represent and infer knowledge

Hierarchy

Multi-level

Multi-scale

Flexible

Spatialorganization

Spatiality

Mobility

Stochasticbehavior

Stochasticity

Semi-permeablecompartments

Encapsulation

Selectivecommunication

Temporalorganization

Dynamic simulations

Biological complexity sets challengingrequirements for computational models

28

/ 85

Models for knowledge integration

Hybridity

Different information

Formalism compatibility or uniformity

Compositionality

Model construction

by composition

Consistency

Leveraging available knowledge and information requires to adapt to its current form.

29

/ 85

Models for knowledge integration

Hybridity

Different information

Formalism compatibility or uniformity

Compositionality

Model construction

by composition

Consistency

Multi-level and hybrid models for systems biologytend combine

differentformalisms with problem-specific

strategies.

30

/ 85

Towards real interdisciplinarity

Jensenius, 2012

Interdisciplinarity requires generality and formalism uniformity.

"Integrating knowledge and methods from differentdisciplines, using a real

synthesis of approaches, towards the creation of a

unity of intellectualframeworks beyond the

disciplinary perspectives"

31

/ 85

Models to represent and exchange knowledge

Findable

For humans

For machines

Accessible

Standardisedcommunication

protocols

Available metadata

Interoperable

Formal, accessible and

shared language

Qualified references to

other (meta)data

Reusable

Rich and relevant

descriptions

Domain-relevant community standards

Wilkinson, 2016

Representations of biological systems need to be F.A.I.R.

32

/ 85

5. A diversity-aware computational framework for systems biology

33

/ 85

The Framework

The Framework includes a

modeling approach

and a description

language to make models accessible

to the entiresystems biology

community.

34

/ 85

Petri Nets

Enabling Delay Firing

Petri Nets arevisual,

mathematicallydefined and executable.

Discrete quantities of resources

Transformation of resources(Transcription, protein activation...)

States of resources(Protein A, protein B, gene X, ...)

35

/ 85

Petri Nets

Enabling Delay Firing

Petri Nets arevisual, mathematically

defined and executable.Expressingtiming and

stochasticity.

Tokens: discrete quantities of resources.

Places: states of resources.

Transitions: transformation of resources.

36

/ 85

Petri Nets

Enabling Delay Firing

Petri Nets arevisual,

mathematicallydefined and executable.Expressingtiming, and

stochasticity.

37

/ 85

Petri Nets

Enabling Delay Firing

Petri Nets arevisual,

mathematicallydefined and executable.Expressingtiming, and

stochasticity.

38

/ 85

Nets-within-nets

NWNs express recursively nested

levels, encapsulation and

selective

communication.

Supportingdynamic

hierarchies and objects.

39

/ 85

Nets-within-nets

Bardini et al, 2018

NWNs express recursively nested

levels, encapsulation and

selective

communication.

Supportingdynamic

hierarchies and objects.

...

40

/ 85

NWNs communication styles

Bardini et al, 2018

Communication channels read or write information

or resources withinor across levels.

41

/ 85Noble, 2015 (modified from Waddington, 1957)Adapted from Salazar et al, 2003

The goal is to model spatiality-mediated localinteractions at

different levels andthe dynamic

hierarchy of regulations they

belong to...

Combined basic mechanisms make a complex landscape

42

/ 85

...to simulate

morphogenetic

pattern formation, considering

architectural and

phenotype

complexity.

Morphogenetic patterns emerge

43

/ 85

A multi-level and multi-context modeling approach

Interactive Spatial Grid Differentiative Landscape

Biological Entities (Cells)

Making the regulation landscape

explicit, mediating

functionalinteractions...

44

/ 85

A multi-level and multi-context modeling approach

Interactive Spatial Grid

Biological Entities (Cells)

...over multiple system levels.

Biological Entities (Tissues)

45

/ 85

A multi-level and multi-context modeling approach

Interactive Spatial Grid

Biological Entities (Cells)

...over multiple system levels.

Biological Entities (Tissues)

Biological Entities (Organs)

46

/ 85

Basic building blocks for local interactions

Interactive Spatial Grid Differentiative Landscape

Biological Entities (Cells)

Contexts havespecific biological

semantics, and are consistent with each

other. Sample building blocks

provide examples of the modeled basic

mechanisms.47

/ 85

Cells building blocks - enzymatic reaction

Black tokens: discrete quantity of molecules.

Places: molecular speciesand states.

Transitions: bioprocesses.

48

/ 85

Cells building blocks - signal sending

Black tokens: discrete quantity of signal in the Cell model.

Colored tokens: discrete quantity of signal in the ISG, and their identity.

Communication channel: passage of the signal acrossthe cell membrane (writing resources to the ISG).

49

/ 85

ISG building blocks – molecular flowColored tokens: discrete quantity of molecules, and their identity.

Net tokens: Cell models.

Places: sub-portions of space.

Transitions: spatiality-mediated interactions between Cell models living in neighboring sub-portions of space.

50

/ 85

ISG building blocks –neighbor communication

Black tokens: discrete quantity of signal in a Cell model.

Colored tokens: discrete quantity of signal with theiridentity in the ISG.

Net tokens: Cell models.

Communication channels: passage of the signal acrossthe cell membrane (writing and reading resources to and from the ISG).

Proximity in the ISG enables interaction,

actualizing the neighborhood relation

51

/ 85

DL building blocks –differentiative stepBlack tokens: discrete quantity of Marker moleculesin Cell models.

Net tokens: Cell models.Places: phenotypes or states.

Transitions: Checkpoints evaluating phenotypechanges by Markers and moving Cells accordingly.

Communication channel: access to Marker by DL (reading information from the Cell).

52

/ 85

Simulations

Model architecture Initial marking

Outputs• Dynamic tracking

of Markers• Pattern

prediction:• Architecture• Phenotypes

Stochastic simulationsexpress process, architectural and

phenotypecomplexity.

53

/ 85

Application examples

54

/ 85

VPC specification in C. Elegans55

/ 85

VPC specification in C. Elegans

Cell model

56

/ 85

VPC specification in C. Elegans

The DL computes cellfates from Marker evolving marking

Cell model

57

/ 85

VPC specification in C. Elegans

Simulations output Markers traces

and a 1D spatialpattern of cell fates

58

/ 85

Antibiotic resistance in the microbiota

Flexibility of the approach: no ISG and two "DL" levels

59

/ 85

Antibiotic resistance in the microbiota

Antibiotic administration

Antibiotic administration + bacterial reintegration

Simulating different clinicalprotocols

60

/ 85

Antibiotic administration

Antibiotic administration + bacterial reintegration

Antibiotic resistance in the microbiota

Model composition and information integration

61

/ 85

Synthetic biology, from modules to systems

Models can help synthetic designs by handling system complexity

62

/ 85

F.A.I.R. enough?

63

/ 85

The modeling approach...64

/ 85

...but VPC specification looks more like this65

/ 85

...or like this

66

/ 85

Abstraction makes things look simpler...67

/ 85

...than they look like in practice...68

/ 8569

/ 85

Findable

For humans

For machines

Accessible

Standardised communiction

protocols

Available metadata

Interoperable

Formal, accessible and

shared language

Qualified references to

other (meta)data

Reusable

Rich and relevant descriptions

Domain-relevant community standards

Is the modeling approach alone F.A.I.R. enough?70

/ 85

Findable

For humans

For machines

Accessible

Standardised communiction

protocols

Available metadata

Interoperable

Formal, accessible and

shared language

Qualified references to

other (meta)data

Reusable

Rich and relevant descriptions

Domain-relevant community standards

...

The framework needs to be relevant for the entire community71

/ 85

Biology System Description Language

72

/ 85

BiSDLgeneric

template

High-level BiSDLmodel descriptionsfollow a generictemplate

73

/ 85

BiSDLgeneric

template

Structural

descriptionscover functionalmodules

Behavioral

descriptionscover theirfunctionalrelationships

74

/ 85

BiSDLdescriptions

75

/ 85

Human readability

Machine-readiness

BiSDLdescriptions

76

/ 85

Findablereferences

Human readability

Machine-readiness

BiSDLdescriptions

77

/ 85

Modules to be shared in libraries

BiSDLdescriptions

Findablereferences

Human readability

Machine-readiness

78

/ 85

Flow

From high-levelBiSDL descriptionsto NWNs models and simulations

79

/ 85

Flow

Computational biologists encode knowledge intomodules, experimental biologists reuse them.

80

/ 85

6. Future perspectives

81

/ 85

Necessary improvements

Modeling

approachHybrid simulations of molecular diffusion

Automated and data-driven model construction

Flow usability BiSDL visual language

Populate BiSDL libraries

Test with diverse user base

Industrial

applications Knowledge management and automation for biotech and pharmaindustries

82

/ 85

First steps

Modeling

approachEfficient simulation of morphogengradient formation

scRNA-seq data analysis of cerebellarastrocytes heterogeneity in development

Flow

usability Bioinformatic pipelines for biologists to characterize nutraceutical agrifoodproducts

Industrial

applications Computational method to improve cell culture processes - patent pending

OR

83

/ 85

1. Bardini, Roberta; Benso, Alfredo; Di Carlo, Stefano; Politano, Gianfranco; Savino, Alessandro; Using nets-within-nets for

modeling differentiating cells in the epigenetic landscape, International Conference on Bioinformatics and Biomedical Engineering, 315-321, 2016, Springer

2. Bardini, Roberta; Di Carlo, Stefano; Politano, Gianfranco; Benso, Alfredo; Modeling antibiotic resistance in the microbiota

using multi-level Petri Nets, BMC systems biology, 12, 6, 108, 2018, BioMed Central

3. Bardini, Roberta; Politano, Gianfranco; Benso, Alfredo; Di Carlo, Stefano; Computational Tools for Applying Multi-level

Models to Synthetic Biology, Synthetic Biology, 95-112, 2018, Springer

4. Muggianu, F; Benso, Alfredo; Bardini, Roberta; Hu, E; Politano, Gianfranco; Di Carlo, Stefano; Modeling biological

complexity using Biology System Description Language (BiSDL), 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 713-717, 2018, IEEE

5. Bardini, Roberta; Politano, Gianfranco; Benso, Alfredo; Di Carlo, Stefano; Using multi-level petri nets models to simulate

microbiota resistance to antibiotics, 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 128-133, 2017, IEEE

6. Bardini, R; Politano, G; Benso, A; Di Carlo, S; Multi-level and hybrid modelling approaches for systems biology, Computational and structural biotechnology journal, 15, 396-402, 2017, Elsevier

7. Distasi, Carla; Dionisi, Marianna; Ruffinatti, Federico Alessandro; Gilardino, Alessandra; Bardini, Roberta; Antoniotti, Susanna; Catalano, Federico; Bassino, Eleonora; Munaron, Luca; Martra, Gianmario; The interaction of SiO2 nanoparticles

with the neuronal cell membrane: activation of ionic channels and calcium influx, Nanomedicine, 14, 5, 575-594, 2019, Future Medicine

• Journal paper extension of Modeling biological complexity using Biology System Description Language (BiSDL)

conference paper (to be submitted in 2019)

• Journal paper centered on the modeling approach and VPC specification example (to be submitted in 2019)84

/ 85

Thank you

85