RETREAT GRBIO 2021

Presentació Metodològica sobre Integració de Dades en Bioinformàtica i Biomedicina. Estat de l'Art i Problemes Oberts

• Un problema que ens preocupa des de fa més d'una dècada és: "Quina és la millor forma de fer d'integrar el màxim d'informació possible per respondre una pregunta científica (sigui el que sigui això)".
• Amb els anys no l'hem respost (o només en part) però hem anat veient com aquesta pregunta es podia reformular de moltes maneres en situacions ben diverses.
• En aquesta xerrada farem una revisió d'algunes aproximacions a la pregunta i a la respostes, tan des d'una òptica general com a través d'alguns dels projectes en què ens veiem implicats com SAMANTHA, VEIS o IMPACT. En concret tocarem dos grans temes.
• D'una banda parlarem del que anomenem "anàlisi integrativa de dades" que, tan en el camp de les òmiques com no òmiques, adquireix molta rellevància amb la creixent disponibilitat de dades de moltes fonts. Revisarem algunes aproximacions, principalment des de l'òptica de l'Anàlisi Multivariant i comentarem alguns problemes oberts en els que voldríem o estem treballant.
• D'altra banda parlarem d'un problema interessant i delicat: com integrar dades d'històries clíniques amb altres -per exemple genòmiques- per a la medicina personalitzada. El desig de crear grans cohorts nacionals i multinacionals d'aquestes característiques porta a importants conflictes tècnics i legals que es miren de resoldre mitjançant diverses aproximacions com la creació de protocols d'intercanvi de dades (el llenguatge OMOP al projecte EHDEN) o la introducció de tècniques d'anàlisi federades que prometen ser una de les claus per a la introducció de la IA en aquest àmbit.

Presentació Metodològica sobre Multistate models

Multi-state models (MSM) dynamically analyze patients' shifts from one disease state to another. They provide an estimation of the transition probabilities as a function of time and on which factors these depend. States of the disease course of COVID-19 hospitalized patients would be, for example, Ward, ICU, discharge, and death.

Does this text sound familiar to you? There it is, it's the one from the first infographic and at GRBIO we are still interested on it, now as a central part of our DIVINE project. In this presentation we will explain what DIVINE is, how we apply MSM and which are the most relevant concepts, assumptions and functions to work with MSM.

TALLER: Fit of multi-state models with R

In this workshop, we will see how to fit multi-state models in R with the mstate package, how to interpret the output, and how to graphically represent the results. For the illustration, we will use the data set of the hospitalized COVID-19 patients in the HM hospitals during the first wave of the COVID-19 pandemic. 

Bioequivalence trials

When the patent on a novel drug expires, drug companies (sponsors) that want to manufacture a copy drug can apply to the regulatory agencies to sell a generic version of the drug. The development of a generic drug are based on the evaluation of the bioequivalence, assessed only in some tens of healthy volunteers. When generic drugs enter the market, the supply extends and drug prices fall making easier access to consumers.

Taller de Mètodes Multivariants pe a la Integració de Dades Òmiques (o no)

El taller s'inspirarà en el capítol 11 del llibre obert "Computational Genomics with R" i, enllaçant amb la primera part de la xerrada anterior practicarem alguns mètodes clàssics i moderns que ens permeten combinar dades de tipus diferents per tal d'aconseguir una explicació més complerta de la que ens oferiria cadascuna de les dades parcials per separat. Ho practicarem amb dades òmiques de càncer que ens descarregarem del TCGA i, "if time permits" veurem com muntar un pipeline complert d'anàlisi amb el paquet "targets".

New kid on the GRBIO block. What will Dani be up in the next 2 years?

As this is my first GRBIO retreat and I might be quite a stranger to quite a few GRBIO members, I will briefly be outlining the lines of future research (out of the GRBIO research projects) I will be working on in the next two years. This mainly will include my research collaboration with Sant Joan de Déu Foundation (CIBERSAM research group) and the clustering model research group of the School of Mathematics and Statistics at the Victoria University of Wellington.

From Artificial Neural Network to Deep Learning

Artificial neural networks is a type of classical algorithm that began to develop in the 50s of the last century. It has some basic elements that give it a very versatile character. It has been in recent years, with an increase in the computing power of computers, when its potential has been harnessed from Deep Learning. In this presentation we will talk about all this and will show some applications in Deep Learning.

The Sorensen-Dice dissimilarity. A journey from ecosystems to gene lists comparison

Genomics (and “omics” in general) provides huge amounts of data, frequently as lists of gene identifiers (e.g., those expressed in a specific disease). A way to biological interpretation of these lists is given by the Gene Ontology, GO, a hierarchic system (from more general to more specific) of biological concepts, designated as GO items or nodes. A gene is “annotated” in a GO item if this concept is applicable to it (e.g., if the gene participates in “membrane transportation”). A GO item is known as “enriched” in a given list if it is overrepresented in the list (overly annotated, more than could be expected by chance). A way to measure biological similarity or dissimilarity between two gene lists could rely on how many enriched GO items they share. Substituting “gene” by “species”, “gene list” by “ecosystem” and “enriched” by “present at”, there are curious coincidences waiting to be exploited.

Sports analytics: overview, R packages, and applications on survival studies

In this talk, we will provide an overview of the role of statistics in the world of sports by presenting some key aspects about the types of data and asking why statistics are increasingly necessary in this field. Likewise, we will present a list of R packages that can be useful on this topic. Finally, we will show how survival analysis can be applied to sports analytics by means of two examples from basketball and Australian rugby.