forked from bastienchassagnol/COPASI_Team216_Lo2016
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathhackathon_keynotes.qmd
128 lines (83 loc) · 5.4 KB
/
hackathon_keynotes.qmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
---
title: "BioModels Hackathon"
author: "bastien"
format: html
---
# Mathematical and Reproducible Models of Human Diseases (BioQuant)
- Make models that can be interpreted, reused and standardised
- conversion of models into useful *kinetic models*, using software Copasi
## Kinetic model
- Biochemical models at the origin.
- Chemical substances converted into other substances, reactions illustrated by arrows.
- *Compartments*: metabolites can be transported from one place to another. Have a volume.
- *Chemical species*: modelled as concentration, *concentrations* are the variables of interest. Can be extended to cell types, biological species, ... Anything that can be quantified as a number, at a certain pace.
- Most complex thing is a *kinetic reaction*, with a given reaction *rate*.
## SBML format, as a computer and semantic representation
- Graphical representation, or matrices of species/variables - reactions/processes.
- Objective of the workshop: convert existing models into the conventional SBML format.
- Model is then converted as a list of ODEs, with simulation:
### General Principle:
For a general reaction of the form,
$$
aA + bB \rightarrow P
$$
the rate law can be written as:
$$
\text{Rate} = k [A]^a [B]^b
$$
Where $a$ and $b$ are the *stoichiometric coefficients*.
In summary, stoichiometric coefficients greater than 1 lead to higher powers in the rate law expression because they reflect the multiplicative nature of interactions between multiple reactant molecules.
# Virtual Patient Engine (VPE)
- A *foundational model* to predict drug targets
- Drug cycle development:
1. Drug Discovery: Target identification, Lead discovery and Lead Optimisation
- Build *biological twins*
- group patients into *responders* and *non-responders* (even super responders, and super non-responders)
- major challenge: multi-scale and multi-system levels
# BioModels: a repository of FAIR and reproducible collection of models
1. Many different modelling approaches: Boolean, ODEs, agent-based, FBA, ...
2. Provide the format (such as SBML)
3. Approve the model
## Curation:
biochemical network -> proper mathematical representation -> computer readable format -> ensure model simulations with Copasi match outputs and model annotation (ontologies)
**BioModels** hosts more than 1,000 models, one of the most popular repository for storing models
Five tabs: overview, files, history (versions), components (the SBML model itself) and curation (verify simulation of the model matches with the figures from the paper) + external links to other databases
- Difference between *replicability(repeatability)* (same exact code) and *reproducibility* (just have the figure, and the inputs)
- Can be considered reproducible if at least one simulation figure can be reproduced, minor deviations are acceptable if they do not change the scientific conclusions
- Half of the models can be reproduced (requirement: equations are provided), 1/3 can not be reproduced at all, even with scientists' reports. Only 1/3 of the researchers could be contacted. Number of models is increasing, but the ratio of non-reproducible papers remains the same.
- reasons for non-reproducibility: missing parameters, initial conditions, structural inconsistency and even, in some cases, the reasons remain unclear.
- How to overcome this challenge? Comply with the reproducibility scorecard (models with score above 4 over 8 are usually reproducible)
# Hackathon Objective:
- Pick up a model from the collection
- Re-encode the model equations in COPASI, and reproduces at least one figure from the manuscript. Does not specifically need to use COPASI, but the one with the most support.
- Export SBML and curate it
# Copasi tutorials
- Depending on the screen size, may require to install it locally.
- constant influx: nothing -> A
- all models are usually composed with some inputs and outputs.
# Code Ocean product demo platform (Youtube Channel in case)
- environment: flex default or dedicated machines
- straightforward deployment
- Jupyter Notebook
- Pipeline is composed of several capsule, with drag and drop system, and custom data // Code is versionned
- Papers are available on Data
- Focused on COPASI, ok as long as the final output is SBML.
- = ->, respectively reversible and irreversible reactions, * to add the stochiometric coefficients
- Output: `B + C ->`
- add to set the quantity unit concentration, does not really change the modelling, but relevant for the curation. Includes as well unit times
- Mathematical equations:
- differential equations: Michaelis-Menten equation can be used to include further equations
-
- Jury evaluation: are we able to reproduce the models? Major objective: improve BioModels repository. Model annotation, from tomorrow, is also relevant.
# Annotation
- additional metadata information provided in addition to the core of the model (structural, mathematical and quantitative elements)
- standard guidelines for biochemical models are MIRIAM
## Reasons
- unambiguous identification of model components
- add semantic layers (better biological understanding)
## What do annotations look like?
- model element (species, reaction or event he whole model) -> qualified ontology (is, hasTaxon, ...) -> cross-reference to an article (CHEBI, Resource identifier)
- types of qualifiers (report to [Identifiers](https://identifiers.org/)):
- bqbiol
## Copasi model
- [Ontology search](https://www.ebi.ac.uk/ols4)