The dissertation aims to discover biological markers of mental illness from functional neuroimaging data. We explored two primary research areas, addressed the following fundamental limitations, and made significant contributions. The first area is the characterization of brain networks in functional magnetic resonance imaging (fMRI), where we answer three key research questions. Firstly, we developed Autolabeller, an automated method enhancing the identification of intrinsic connectivity networks in fMRI data, improving accuracy and reproducibility while mitigating noise and artifacts. Secondly, we adapted methods for reconstructing subject-specific networks in schizophrenia patients and controls using spatiotemporal regression and information-guided independent component analysis. Lastly, we identified optimal cluster validity indexes, enhanced dynamic functional network connectivity analysis, and overcame existing ambiguities in brain network identification. In identifying biomarkers for brain disorders, we pursued four key research questions. Firstly, we applied information theory to pinpoint biomarkers specific to schizophrenia using the dynamic functional domain connectivity framework. The analysis revealed heightened entropy and diminished cross-domain mutual information in select domain pairs, indicating increased uncertainty and reduced interdependence among brain functions in schizophrenia patients. Secondly, we harnessed machine learning techniques to identify image-derived biomarkers for schizophrenia, showcasing the sensitivity of adaptive brain networks in achieving consistent classification results across independent datasets. Additionally, we introduced a biologically-based classification algorithm within the neuromark framework to predict treatment response accurately, particularly in bipolar disorder patients presenting with predominant depressive symptoms. This approach holds promise for uncovering biomarkers linked to medication class responses within mood disorders. Lastly, we explored topological data analysis for biomarker identification across brain disorders, revealing reduced dynamism in schizophrenia patients.