2026 ECNP Congress

PS02-2161

Transdiagnostic resting-state fMRI biotypes in early psychosis and depression predict cognitive deficits and functional outcomes

Nikos Fabian Diederichs — M.O. Buciuman, C.S. Vetter, C. Weyer, P. Zhutovsky, S. Tovar Perdomo, A. Khuntia, Y. Milaneschi, D. Popovic, A. Ruef, D.B. Dwyer, K. Chisholm, L. Kambeitz-Ilankovic, L.A. Antonucci, T. Lichtenstein, M. Rosen, S. Ruhrmann, J. Kambeitz, A. Riecher-Rössler, R. Upthegrove, R.K.R. Salokangas, J. Hietala, C. Pantelis, R. Lencer, E. Meisenzahl, S.J. Wood, P. Brambilla, S. Borgwardt, A. Bertolino, P. Falkai, N. Koutsouleris, P. Lalousis · LMU University Hospital Munich, Department of Psychiatry and Psychotherapy, Munich, Germany

2026 ECNP Congress

Early psychosis and depression exhibit substantial diagnostic heterogeneity, complicating prognosis and treatment selection. Approximately 63% of depressed patients do not remit with first-line treatment [1], underscoring the need for biologically informed stratification. Previous data-driven studies identified grey matter volume-based transdiagnostic subgroups, consistently revealing an anatomically impaired subtype with poorer functioning [2]. However, resting-state fMRI captures complementary information about large-scale network dynamics. Both psychosis-spectrum and mood disorders show dysregulation of default mode, frontoparietal, and salience networks, yet no study has investigated transdiagnostic rs-fMRI biotypes across early psychosis and depression using semi-supervised clustering. We aimed to identify and characterise rs-fMRI-based transdiagnostic subtypes in clinical high-risk (CHR), recent-onset psychosis (ROP), and recent-onset depression (ROD) from the multi-site PRONIA consortium [3].

Participants were recruited across eight European sites (PRONIA): N=419 healthy controls (HC), N=289 CHR, N=270 ROD, and N=318 ROP (total N=1,296). Resting-state fMRI preprocessing used SPM12 with standard denoising (nuisance regression, Friston-24 motion parameters, despiking, detrending). Thirty probabilistic functional modes (PFMs) were estimated using the PROFUMO variational Bayesian framework across all participants, capturing inter-individual variability without imposing spatial or temporal orthogonality [4]. Mean PFM amplitudes were extracted per participant. A hierarchical Bayesian normative modelling (HBNM) pipeline trained on HC generated individual deviation z-scores, modelling age, sex, head motion, and site as a hierarchical random effect (robust Student-t, 8,000 iterations, 4 chains). HBNM validation confirmed HC site prediction dropped to 1.2x chance (0/28 site pairs distinguishable). Heterogeneity through Discriminative Analysis (HYDRA) semi-supervised clustering identified patient subgroups relative to healthy norms [5]. Optimal cluster number was determined via multi-metric consensus (Adjusted Rand Index [ARI], Silhouette, Calinski-Harabasz, BIC) across 1,000 cross-validation permutations. Subgroups were characterised using PANSS, SANS, BDI, cognitive assessments, polygenic risk scores (PRS), and psychosocial functioning (GAF, GF scales) at baseline and follow-up.

The optimal transdiagnostic solution comprised three subtypes (ARI=0.79, 95% CI [0.78, 0.80]; stability=0.90), robust to feature exclusion (k=3 retained without Feature 30; ARI=0.68). HBNM achieved superior clustering stability over standard Bayesian ComBat harmonisation for the two-subtype comparison (ARI=0.73 vs. 0.65). Top discriminating PFMs mapped to frontoparietal control (|d|=1.42), DMN/precuneus (|d|=1.35), and salience/prefrontal (|d|=1.05) networks. The impaired subtype showed globally reduced PFM amplitudes except for salience and somatomotor network hyperactivation. Robust cognitive differences emerged: the impaired subtype exhibited significantly lower verbal learning and global cognitive ability (FDR-corrected p<.05). Cognitive PRS was the only significant polygenic signal, suggesting a transdiagnostic cognitive vulnerability dimension independent of disorder-specific PRS (depression, schizophrenia: n.s.). Crucially, BDI depression severity was virtually identical between subtypes (25.3 vs. 25.8, p>.7), confirming biological rather than severity-based clustering. Baseline biotype membership predicted persistent poor functioning at 9-18 month follow-up (GAF Symptoms: chi-squared=8.8, p=.003; GF Role: chi-squared=6.4, p=.01).

We identify transdiagnostic rs-fMRI biotypes with distinct cognitive, polygenic, and functional outcome profiles, supporting biologically informed alternatives to categorical diagnosis. HBNM improved clustering stability over standard harmonisation, demonstrating the importance of principled Bayesian site correction in multi-site neuroimaging subtyping. The impaired biotype’s cognitive-genetic signature and longitudinal prognostic value suggest utility for early stratification beyond traditional diagnostic boundaries. External validation in the NESDA cohort (N=301) is underway.

References

  1. Rush, A.J., Trivedi, M.H., Wisniewski, S.R., Nierenberg, A.A., Stewart, J.W., Warden, D., Niederehe, G., Thase, M.E., Lavori, P.W., Lebowitz, B.D., McGrath, P.J., Rosenbaum, J.F., Sackeim, H.A., Kupfer, D.J., Luther, J., Fava, M., 2006. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR*D report. Am. J. Psychiatry 163 (11), 1905-1917.
  2. Lalousis, P.A., Schmaal, L., Wood, S.J., Reniers, R.L.E.P., Barnes, N.M., Chisholm, K., Griffiths, S.L., Stainton, A., Wen, J., Hwang, G., Davatzikos, C., Wenzel, J., Kambeitz-Ilankovic, L., Andreou, C., Bonivento, C., Dannlowski, U., Ferro, A., Lichtenstein, T., Riecher-Rössler, A., Romer, G., Rosen, M., Bertolino, A., Borgwardt, S., Brambilla, P., Kambeitz, J., Lencer, R., Pantelis, C., Ruhrmann, S., Salokangas, R.K.R., Schultze-Lutter, F., Schmidt, A., Meisenzahl, E., Koutsouleris, N., Dwyer, D., Upthegrove, R., 2022. Neurobiologically based stratification of recent-onset depression and psychosis: identification of two distinct transdiagnostic phenotypes. Biol. Psychiatry 92 (7), 552-562.
  3. Koutsouleris, N., Kambeitz-Ilankovic, L., Ruhrmann, S., Rosen, M., Ruef, A., Dwyer, D.B., Paolini, M., Chisholm, K., Kambeitz, J., Haidl, T., Schmidt, A., Gillam, J., Schultze-Lutter, F., Falkai, P., Reiser, M., Riecher-Rössler, A., Upthegrove, R., Hietala, J., Salokangas, R.K.R., Pantelis, C., Meisenzahl, E., Wood, S.J., Beque, D., Brambilla, P., Borgwardt, S., 2018. Prediction models of functional outcomes for individuals in the clinical high-risk state for psychosis or with recent-onset depression. JAMA Psychiatry 75 (11), 1156-1172.
  4. Harrison, S.J., Woolrich, M.W., Robinson, E.C., Glasser, M.F., Beckmann, C.F., Jenkinson, M., Smith, S.M., 2020. Modelling subject variability in the spatial and temporal characteristics of functional modes. NeuroImage 222, 117226.
  5. Varol, E., Sotiras, A., Davatzikos, C., 2017. HYDRA: Revealing heterogeneity of imaging and genetic patterns through a multiple max-margin discriminative analysis framework. NeuroImage 145, 346-364.