Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Sketchpad Series 

Neurosynth Compose: A platform for transparent and reproducible meta-analyses

Poster Session B - Sunday, March 8, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom
Also presenting in Data Blitz Session 4 - Saturday, March 7, 2026, 10:30 am – 12:00 pm PST, Salon F.

James Kent¹ (), Nicholas Lee², Taylor Salo³, Katherine Bottenhorn⁴, Jerome Dockes⁵, Ross Blair⁶, Thomas Nichols⁷, Angela Laird⁴, Jean-Baptiste Poline², Tal Yarkoni¹, Alejandro De La Vega¹, Grace Robertson⁸, Amy Ramage⁸; ¹University of Texas at Austin, ²McGill University, ³University of Pennsylvania, ⁴Florida International University, ⁵Inria, France, ⁶Stanford University, ⁷University of Oxford, ⁸University of New Hampshire

Neuroimaging meta-analyses are a cornerstone of cognitive neuroscience, enabling synthesis across decades of neuroimaging research. However, the traditional workflow of manually curating and coding thousands of papers to extract activation coordinates is idiosyncratic and time-consuming. To address this, we developed Neurosynth Compose, an open platform that facilitates the creation of meta-analyses with transparency and reproducibility at its core. The platform has automatically extracted reported activation coordinates and key metadata (e.g., participant characteristics, tasks, populations) from over 32,000 neuroimaging studies, supporting both hypothesis-driven and data-driven exploration through AI-assisted curation. With over 2,500 users worldwide, Neurosynth Compose offers a scalable infrastructure for systematic meta-analysis, encouraging the collaborative reuse of existing research. To demonstrate its capabilities, we are conducting an exemplar analysis comparing the neural bases of American Sign Language (ASL) and spoken language production. Using studies within Neurosynth Compose and newly added literature, we apply Activation Likelihood Estimation (ALE) and Multilevel Kernel Density Analysis (MKDA) with Monte Carlo-based familywise error correction (5,000 iterations). Comparative analyses (ALE subtraction and MKDA χ² tests) will illustrate cross-method consistency and the platform’s ability to synthesize results reproducibly. Neurosynth Compose lowers the barrier to meta-analysis, increases methodological transparency, and enables the field to build on its vast cumulative knowledge. This open, automated framework advances the collective goal of grounding cognitive neuroscience on reproducible, integrative evidence.

Topic Area: METHODS: Neuroimaging