Experimental assessment of AI-based interactome mapping

L Luke Lambourne A Anupama Yadav Y Yang Wang A Alice Desbuleux D Dae-Kyum Kim F Florent Laval K Kerstin Spirohn-Fitzgerald T Tiziana Cafarelli C Carles Pons I István A. Kovács N Noor Jailkhani S Sadie Schlabach D David De Ridder K Katja Luck V Vladimir V. Botchkarev O Olivia Debnath W Wenting Bian Y Yun Shen Z Zhipeng Yang (Engineering Research Center of Coptis Development and Utilization (Ministry of Education), College of Pharmaceutical Sciences, Southwest University) M Miles W. Mee M Mohamed Helmy Y Yves Jacob (Unité de Génétique Moléculaire des Virus à ARN, CNRS UMR 3569, Département Virologie, Institut Pasteur) I Irma Lemmens T Thomas Rolland G Gregory G. McClain A Atina G. Coté M Marinella Gebbia N Nishka Kishore J Jennifer J. Knapp J Joseph C. Mellor G Gonen Memisoglu J Jüri Reimand J Jan Tavernier M Michael E. Cusick Q Quan Zhong P Patrick Aloy T Tong Hao B Benoit Charloteaux F Frederick P. Roth J Javier De Las Rivas P Pascal Falter-Braun D David E. Hill M Michael A. Calderwood J Jean-Claude Twizere M Marc Vidal

Abstract

Abstract Genotype-phenotype relationships are mediated through intricate networks of physical and functional interactions among macromolecules. Knowledge of the interactome is vital to understand and model genetics and cellular biology. Recent advances in accurately predicting tertiary protein structures using artificial intelligence (AI) approaches such as AlphaFold 1 have revived the vision that the protein-protein interactome might be fully predictable through computational modeling of quaternary structures. Here we present a comprehensive experimental framework to systematically assess the impact of AI-driven interactome predictions for yeast 2 and human 3 . We find that the quality of high-confidence predictions is on par with established experimental approaches. However, in proteome-wide screening, the tested AI approaches underperform in the discovery of strictly novel protein-protein interactions (PPIs) compared to experimental reference interactome maps. In particular, the yeast interactome map described here identifies >40-fold more novel PPIs than its AI counterpart. Strikingly, AlphaFold provides structural models for a substantial number of experimentally identified PPIs missed by the virtual screens. Our results suggest that, at this stage, the main contribution of AI predictions is to provide quaternary structure models for experimentally identified PPIs.

Article Details

Volume / Issue Vol. 17, Issue 1
Published April 04, 2026
ISSN 2041-1723
Publisher Nature Portfolio

Journal Info

Nature Communications

Nature Portfolio

ISSN: 2041-1723 Open Access Life Sciences

Authors (45)

L

Luke Lambourne

A

Anupama Yadav

Y

Yang Wang

A

Alice Desbuleux

D

Dae-Kyum Kim

F

Florent Laval

K

Kerstin Spirohn-Fitzgerald

T

Tiziana Cafarelli

C

Carles Pons

I

István A. Kovács

N

Noor Jailkhani

S

Sadie Schlabach

D

David De Ridder

K

Katja Luck

V

Vladimir V. Botchkarev

O

Olivia Debnath

W

Wenting Bian

Y

Yun Shen

Z

Zhipeng Yang

Engineering Research Center of Coptis Development and Utilization (Ministry of Education), College of Pharmaceutical Sciences, Southwest University

M

Miles W. Mee

M

Mohamed Helmy

Y

Yves Jacob

Unité de Génétique Moléculaire des Virus à ARN, CNRS UMR 3569, Département Virologie, Institut Pasteur

I

Irma Lemmens

T

Thomas Rolland

G

Gregory G. McClain

A

Atina G. Coté

M

Marinella Gebbia

N

Nishka Kishore

J

Jennifer J. Knapp

J

Joseph C. Mellor

G

Gonen Memisoglu

J

Jüri Reimand

J

Jan Tavernier

M

Michael E. Cusick

Q

Quan Zhong

P

Patrick Aloy

T

Tong Hao

B

Benoit Charloteaux

F

Frederick P. Roth

J

Javier De Las Rivas

P

Pascal Falter-Braun

D

David E. Hill

M

Michael A. Calderwood

J

Jean-Claude Twizere

M

Marc Vidal