Valentina Giunchiglia

Postdoctoral Researcher · Genentech & Stanford University

Valentina Giunchiglia

About

I am a postdoctoral researcher at Genentech and Stanford University, supervised by Jason Vander Heiden and Tina Hernandez-Boussard. I build multimodal AI models that integrate molecular data with clinical and electronic health records to better understand patients’ immune phenotypes and predict longitudinal patient trajectories.

I completed my PhD jointly at Imperial College London and Harvard University, supported by an MRC multi-scale fellowship and supervised by Marinka Zitnik, Adam Hampshire, and Richard Nicholas. My research asks how precise molecular changes map to specific clinical and patient-level phenotypes, with the goal of improving decisions about when to treat a patient, which medication to use, whom to recruit into clinical trials, and which drugs to prioritise for development.

I have authored more than 35 publications, including over ten as first or joint-first author, in venues including Cell, Nature Communications, The New England Journal of Medicine, Nature Medicine, npj Digital Medicine, and ICLR.

Education

Research

To map molecular signals to clinical phenotypes — and ultimately bring these models into the clinic — I work in two main directions. The first focuses on developing and validating AI models that learn multiscale interactions and connect molecular and cellular variation to disease phenotypes, at both the cohort and individual-patient level. The second focuses on modelling clinical phenotypes with greater precision, moving beyond broad diagnostic labels such as “Alzheimer’s disease” towards detailed clinical fingerprints that capture the heterogeneity of individual patients.

My long-term goal is to integrate these two directions into a unified framework that achieves precision along both dimensions: the molecular and the clinical. I am particularly interested in developing models that can ultimately be deployed in clinical settings and have a tangible translational impact. To build towards this, my work sits at the intersection of three closely connected components: model development, rigorous validation, and model explanation, which I consider essential for developing AI systems that can be trusted and used in clinical practice.

Research questions I’m interested in
  • How can we build longitudinal, multimodal representations of patients that capture how molecular and clinical states evolve over time and predict future disease trajectories?
  • How can we model the effect of interventions on patient trajectories?Can we predict how a patient’s future state would change following a treatment, disease event, or behavioural intervention, and compare multiple plausible trajectories under different interventions and at different timepoints?
  • When is the optimal time to intervene?Can longitudinal models identify when a specific treatment, molecular target, or clinical intervention is most likely to alter a patient’s trajectory?
  • Which data are necessary, and when are they necessary?At each point in a patient’s trajectory, which molecular, clinical, imaging, or other measurements provide the most useful information for prediction or decision-making? Can we determine when additional data collection is unlikely to meaningfully improve a clinical decision?
  • Which molecular signals actually explain variation in clinical phenotypes?Can models move beyond association to identify the molecular pathways, cell states, or targets that contribute to specific patient-level clinical phenotypes and trajectories?
  • How can we rigorously determine whether learned representations are biologically meaningful and clinically relevant?Do model representations recover known biology, generalise across cohorts and healthcare systems, and predict clinically meaningful outcomes in unseen patients?
  • How much information is enough for clinical use?What is the minimum set of measurements required to achieve reliable predictions, and how does model performance change when expensive, invasive, or difficult-to-obtain modalities are removed?
  • Can models decide when they have enough evidence to act?Can an AI system quantify when a prediction or recommendation is sufficiently reliable to support a clinical decision, when it should defer to a clinician, and when additional information should be collected?
  • How can models actively guide what should be measured next?Given the information already available for a patient, can the model identify the next most informative test, molecular assay, clinical assessment, or follow-up timepoint?
  • Can we model uncertainty over multiple plausible patient futures rather than predicting a single trajectory?How can we represent heterogeneous possible outcomes and distinguish uncertainty caused by incomplete information from genuine biological variability?
  • How transferable are molecular–clinical relationships across patients, diseases, and populations?Which learned mechanisms are patient-specific, which generalise across subgroups, and which remain stable across cohorts, institutions, and disease contexts?

Software & datasets

ModelCASCADE

Context-aware single-cell foundation models linking cellular programmes to patient-level disease phenotypes, with disease-specific models for Alzheimer’s, autism, thyroid and lung.

Website Code Models
ModelProCyon

Multimodal foundation model for protein phenotypes across molecular function, therapeutic mechanism, disease association, protein domains and interactions.

Paper Website Code Models
MethodIDoCT

Iterative Decomposition of Cognitive Tasks — disentangles cognitive ability from motor and device speed in large-scale online and computerised cognitive assessments.

Code Paper · npj Digital Medicine Paper · Imaging Neuroscience
MethodEXPASS

Explanation-directed message passing for graph neural networks — aggregates only the nodes and edges that a GNN explanation method identifies as important.

Code Paper · LoG 2022
DatasetUK Biobank cognitive phenotypes↳ from IDoCT

Domain-specific cognitive ability and visuo-motor speed scores generated by IDoCT for the UK Biobank imaging cohort and returned to UK Biobank as derived phenotypes (the dataset used in the Imaging Neuroscience paper).

UK Biobank data Paper
DatasetMultiple sclerosis target RNA-seq↳ from ProCyon

Bulk RNA-seq generated to test ProCyon-nominated multiple sclerosis targets, from the Multiple Sclerosis Journal analysis. Public release on GEO in preparation.

GEO · coming soon MS analysis ProCyon paper

Publications

* equal contribution  ·  Google Scholar

2026

  1. CASCADE: context-aware single-cell modelling links cellular programmes to patient-level disease phenotypes

    V. Giunchiglia*, et al.

    In submission

    WebsiteCodeModels

  2. Digital cognitive phenotyping for differential diagnosis and monitoring in neurological conditions

    M. Del Giovane, V. Giunchiglia, M. C. B. David, M. A. Kolanko, W. R. Trender, et al.

    Annals of Clinical and Translational Neurology

    Paper

  3. Medea: an omics AI agent for therapeutic discovery

    P. Sui, M. M. Li, S. Gao, W. Shen, V. Giunchiglia, et al.

    bioRxiv

    Paper

  4. Validation of the Comprehensive Online Sleep Monitoring Scale (COSMOS) in a large population sample

    L. Rida, K. Ioannidis, S. R. Chamberlain, J. E. Grant, P. Hellyer, V. Giunchiglia, et al.

    Frontiers in Psychology

    Paper

2025

  1. ProCyon: a multimodal foundation model for protein phenotypes

    O. Queen*, Y. Huang*, R. Calef*, V. Giunchiglia*, et al.

    bioRxiv · under review

    PaperWebsiteCodeModels

  2. Large-scale online assessment uncovers a distinct multiple sclerosis subtype with selective cognitive impairment

    A. Lerede, A. Moura, V. Giunchiglia, E. Carta, et al.

    Nature Communications

    Paper

  3. Post-hospitalisation COVID-19 cognitive deficits at one year are global and associated with elevated brain injury markers and grey-matter volume reduction

    G. K. Wood, B. F. Sargent, Z. U. A. Ahmad, K. Tharmaratnam, …, V. Giunchiglia, et al.

    Nature Medicine

    Paper

  4. Development and validation of the IC3: an online remote assessment technology for deep phenotyping and monitoring of cognitive impairment after stroke

    D. C. Gruia, V. Giunchiglia, A. Coghlan, S. Brook, et al.

    Assessment

    Paper

  5. Mitigating the impact of motor impairment on self-administered digital tests in patients with neurological disorders

    D. C. Gruia, V. Giunchiglia, et al.

    eClinicalMedicine

    Paper

  6. Remote cognitive tests predict neurodegenerative biomarkers in the Insight 46 cohort

    M. Del Giovane, V. Giunchiglia, Z. Cai, et al.

    Alzheimer’s & Dementia

    Paper

  7. Online46: online cognitive assessments in elderly cohorts — the British 1946 birth cohort case study

    Z. Cai*, V. Giunchiglia*, R. Street*, M. Del Giovane, et al.

    Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring

    Paper

  8. Exploring novel biologically plausible targets in multiple sclerosis using the multimodal ProCyon foundation model

    O. Queen*, V. Giunchiglia*, G. Abbadessa, et al.

    Multiple Sclerosis Journal

    Paper

  9. KGARevion: an AI agent for knowledge-intensive biomedical QA

    X. Su, Y. Wang, S. Gao, X. Liu, V. Giunchiglia, et al.

    International Conference on Learning Representations (ICLR)

    Paper

  10. Application of Whisper in clinical practice: post-stroke speech assessment during a naming task

    M. Davudova*, Z. Cai*, V. Giunchiglia, D. C. Gruia, et al.

    NeurIPS GenAI Workshop

    Paper

  11. Persistent multi-system impairments detected by wearable monitoring in long-COVID cases reporting fatigue up to three years post-infection

    M. Orini, C. Stuart, A. Jamieson, A. Fernandez-Sanles, Y. Ranjan, …, V. Giunchiglia, et al.

    Preprint (Research Square)

    Paper

  12. Meta-analysis of 22,710 human microbiome metagenomes

    P. Manghi, G. Antonello, L. Schiffer, D. Golzato, …, V. Giunchiglia, et al.

    Nature Communications

  13. Neurological and psychiatric complications of COVID-19 (COVID-CNS): a national cohort study

    R. S. K. Shil, A. Seed, N. E. Franklyn, …, V. Giunchiglia, et al.

    Scientific Reports

  14. Usability and compliance of online cognitive assessments in the British 1946 birth cohort

    V. Giunchiglia, Z. Cai, M. Del Giovane, et al.

    Alzheimer’s Association International Conference · abstract

2024

  1. An iterative approach for estimating domain-specific cognitive abilities from large-scale online cognitive data

    V. Giunchiglia, D. C. Gruia, A. Lerede, W. Trender, et al.

    npj Digital Medicine

    Paper

  2. Neural correlates of cognitive ability and visuo-motor speed: validation of IDoCT on UK Biobank data

    V. Giunchiglia, S. Curtis, S. Smith, N. Allen, A. Hampshire.

    Imaging Neuroscience

    Paper

  3. Empowering biomedical discovery with AI agents

    S. Gao, A. Fang*, Y. Huang*, V. Giunchiglia*, A. Noori*, et al.

    Cell

    Paper

  4. Cognition and memory after Covid-19 in a large community sample

    A. Hampshire, A. Azor, C. Atchison, W. Trender, P. J. Hellyer, V. Giunchiglia, et al.

    New England Journal of Medicine

    Paper

  5. Online cognitive monitoring technology for people with Parkinson’s disease and REM sleep behaviour disorder

    M. Bălăeţ, F. Alhajraf, T. Zerenner, J. Welch, J. Razzaque, C. Lo, V. Giunchiglia, et al.

    npj Digital Medicine

    Paper

  6. Remote cognitive testing in Insight46: relationship to standard cognitive assessments and biomarkers of neurodegeneration

    M. Del Giovane, M. Leoni, V. Giunchiglia, Z. Cai, et al.

    Alzheimer’s & Dementia · abstract

  7. Monitoring vascular cognitive impairment using a scalable digital online cognitive tool

    D. C. Gruia, V. Giunchiglia, A. Coghlan, et al.

    International Journal of Stroke · abstract

  8. Widespread cell loss in the superior frontal gyrus of multiple sclerosis subjects detected using a novel automated cell-counting algorithm

    A. Mahapatra, V. Giunchiglia, et al.

    Multiple Sclerosis Journal · abstract

Earlier work (2023–2020)

2023

  1. The effects of COVID-19 on cognitive performance in a community-based cohort: a COVID Symptom Study Biobank prospective cohort study

    N. J. Cheetham, R. Penfold, V. Giunchiglia, et al.

    eClinicalMedicine

  2. Moving from phenomenological to predictive modelling: progress and pitfalls of modelling brain stimulation in silico

    D. L. Kurtin, V. Giunchiglia, J. Vohryzek, et al.

    NeuroImage

    Paper

  3. Computerised cognitive assessment in patients with traumatic brain injury: an observational study of feasibility and sensitivity relative to established clinical scales

    M. Del Giovane, W. R. Trender, M. Bălăeţ, …, V. Giunchiglia, et al.

    eClinicalMedicine

  4. Computerised cognitive testing and multi-domain structural MRI in idiopathic normal-pressure hydrocephalus and Alzheimer’s disease

    M. Del Giovane, T. D. Parker, M. C. B. David, …, V. Giunchiglia, et al.

    Alzheimer’s & Dementia · abstract

  5. The motor and cognitive components of impaired performance in online cognitive tasks and their association with patient-reported outcomes

    V. Giunchiglia*, A. Lerede*, W. Trender, et al.

    Multiple Sclerosis Journal (ECTRIMS) · abstract

2022

  1. Towards training GNNs using explanation-directed message passing

    V. Giunchiglia*, C. V. Shukla*, G. González, C. Agarwal.

    Learning on Graphs Conference (LoG)

    Code

  2. Adverse childhood experiences and severity levels of inflammation and depression from childhood to young adulthood: a longitudinal cohort study

    E. Iob, R. Lacey, V. Giunchiglia, A. Steptoe.

    Molecular Psychiatry

    Paper

  3. Automated cancer diagnostics via analysis of optical and chemical images by deep and shallow learning

    O. G. Isberg, V. Giunchiglia, J. S. McKenzie, et al.

    Metabolites

    Paper

  4. Policies and politics: an analysis of the public policies aimed at the reorganisation of healthcare delivery during the COVID-19 pandemic

    S. Lewis-Jackson, E. Iob, V. Giunchiglia, et al.

    Caring on the Frontline during COVID-19 (Palgrave Macmillan)

  5. Diffusely abnormal white matter and elevated grey-matter demyelination imply rapid and severe progression in atypical multiple sclerosis

    R. Magliozzi*, V. Giunchiglia*, A. Mensi*, et al.

    Multiple Sclerosis Journal (ECTRIMS) · abstract

  6. An automated data-cleaning approach to remove preparation artefacts from brain histology slide images

    V. Giunchiglia, S. Gentleman, R. Nicholas.

    Movement Disorders · abstract

2020

  1. Precursors for nonlymphoid-tissue Treg cells reside in secondary lymphoid organs and are programmed by the transcription factor BATF

    M. Delacher, C. D. Imbusch, A. Hotz-Wagenblatt, …, V. Giunchiglia, et al.

    Immunity

    Paper

Teaching

Course material is openly available — each item links to its own lecture or folder.

MSc Translational Neuroscience · Imperial College London

Brain sciences

Computational Methods for the Brain Sciences

Lead Teaching Assistant · 2021–2024

An intensive module introducing computational methods for the brain sciences through lectures, workshops and hands-on challenges, spanning programming, big-data and cognitive analysis, neuroimaging, brain connectivity and machine learning. I designed and delivered the material and co-supervised the end-of-module projects.

  1. Introduction to programming — Python and data visualisation
  2. Big-data analysis — COVID-19 and cognitive function
  3. Cognitive analysis — self-harm and cognition
  4. Neuroimaging fundamentals — fMRI and structural MRI
  5. Group-level fMRI analysis
  6. Brain connectivity — graph theory applied to fMRI
  7. Unsupervised machine learning — clustering and pattern discovery
  8. Supervised machine learning — predicting neurological outcomes
All teaching material →
Programming

Python & statistics pre-course

Course designer & lecturer · 2021–2024

A ten-lecture Jupyter-notebook course, with graded exercises and platform setup guides, that brings incoming students up to speed on Python and statistics before the taught modules begin.

  1. Syntax, variables, operators and file handling
  2. Functions, scripting and debugging
  3. Lists, arrays, imports and mathematical operations
  4. Conditional statements
  5. Loops and iteration
  6. Strings and string methods
  7. Dictionaries
  8. DataFrames — creation, cleaning and merging
  9. Data visualisation
  10. Statistics and hypothesis testing
Exercises → All teaching material → Course development (with D. Gruia) →

Data science for biomedical research · Imperial College London

Data science

Biomedical Data Science — Data Science Helper Team

Founder, organiser & lecturer · 2020–2021

I founded a peer-teaching team that ran a series of introductory lectures and coding practicals, bookable one-to-one clinics, and a discussion forum for master’s students.

All materials & schedule →

Bioinformatics

Bioinformatics

Better Bioinformatics Seminar

Organiser · biweekly · 2025–

A biweekly seminar series I organise on rigorous, reproducible bioinformatics practice — every other Tuesday, with invited speakers, recordings and a companion resources page.

Resources → Schedule → Website →
Bioinformatics

Introduction to Bioinformatics — cancer methylome project

Tutor · BSc Molecular Biotechnology, Heidelberg University · 2018–2019

Tutored a group project (documented in R Markdown) on the cancer methylome: identifying differentially methylated regions between disease and healthy haematopoietic samples using Blueprint Epigenome data — from quality control and normalisation of methylation matrices, through PCA and clustering, to differential-methylation testing, logistic regression for disease status, and genomic annotation.

Course project & materials →

Outreach teaching

Machine learning

Machine learning from scratch — GirlsWhoML

Demonstrator · 2021

Tutorials on implementing and applying linear and logistic regression from scratch, for students at BSc, MSc and PhD level.

GirlsWhoML →

Outreach

Places where I have given a talk, presented a poster, or organised a conference or event — drag to spin, and hover or tap a point for details.

  • –Talks
  • –Posters
  • –Events organised
  • –Places