Schizophrenia is a neurodevelopmental disorder characterized by hyperfunctional symptoms such as delusions and hallucinations, but uniquely also several hypofunctional symptoms such as social withdrawal, avolition, and worsened cognitive functioning [1, 2]. The pathophysiology of schizophrenia is still largely unknown, but speculative work centers around abnormalities in neurotransmission and neurochemical balance in the form of abhorrent neurotransmitter activity and neurochemicals such as gamma-aminobutyric acid (GABA) and aspartate [1, 2]. Moreover, schizophrenia’s neurodevelopmental nature complicates pathophysiology, as presentation and/or development itself can depend on environmental factors [1, 2]. With schizophrenia being labeled as a neurodevelopmental disorder as recently as three decades ago, research regarding the heterogeneity of the disorder across demographics is still limited, making it difficult to design individualized medication to alleviate symptoms. 

Prior to advancements in the neurotransmitter hypothesis of schizophrenia, medication was limited to “first generation antipsychotics”, which targeted dopamine D2 receptors in the mesolimbic pathways which decreased positive symptoms but disregarded symptoms unrelated to excitatory cognitive function [3]. Currently, medication for schizophrenia, or second generation antipsychotics, aim to ease symptoms by dually and partially antagonizing dopamine D2 receptors in the mesolimbic pathway and serotonin 5-HT2A receptors in the mesocortical pathway, which increases dopamine helping with social apathy as well as reduces tremors introduced by first generation medication [3, 4]. In some cases, benzodiazepines are prescribed as an adjunct treatment to manage agitation, catatonia, and anxiety. They enhance GABA-A receptor activity which increases neuronal inhibition and creates a calming effect [3, 4]. However, current medication is nonspecific and does not target the core etiology of schizophrenia, evident by the fact that one third of patients are treatment resistant to current antipsychotics [4]. Furthermore, the side effects are detrimental, making it difficult for patients to consistently stay on their medication. Side effects include neuromuscular dysfunction, skin and eye reactions, cardiovascular risks, sedation, increased risk of cognitive decline, and physical dependence [4]. One of the most overlooked symptoms, sexual dysfunction [4], is also one of the most common reasons for patients discontinuing treatment, which is vital for treatment success.

The non-specificity, damaging effects, and low adherence potential of current medications indicate the urgent need for precise and individualized treatment. My project aims to address this by working with fMRI and multi omic data to develop a prototype pharmaceutical that targets the underlying dysfunction instead of solely modulating neurotransmitters. This work can guide future drug development as it will not only highlight the differences in neurodevelopment across demographics but also compare the efficacy of current treatment and treatment that aims to restore balance rather than alter.

fMRI Analysis:

 

Materials:

Google Scholar and various publishers

CONN Toolbox

FSLeyes

 

1. Literature Dive - I began with a literature dive to look into aging, suicidality, treatment resistance, and sleep spindle abnormalities, with a keen focus on epigenetic alterations. Throughout my study, I consulted literature to ensure all assertions were grounded in the current scientific understanding of schizophrenia.
2. Data acquisition - Data of 72 control and 75 patient fMRI scans were obtained from the Centre of Biomedical Research Excellence (COBRE) and downloaded to my computer.
3. Excluded any patients (2, one patient and 1 control) with other psychiatric disorders to avoid comorbidities affecting results.

 

Analysis in the CONN Toolbox

 

1. Loaded samples’ structural and anatomical data files in the CONN toolbox
2. Preprocessed data by realigning images to become standardized, skull-stripped segmented images, normalized data to the standard MNI space, and smoothed using Gaussian kerning available in the CONN toolbox. Utilized Artifact Detection Tools (ART) and principle component filtering of signal from the white matter and cerebral spinal fluid (Compcor). The main goal was to remove motion related artifacts which can severely alter results. 
3. Performed quality assurance checks by examining images by displaying slices with MNI boundaries, ROI segmentation, atlas network parcellations, and covariates like motion to see if they have been eliminated.
4. Denoised samples - rsfMRI is very susceptible to linear drifts and motion - removed any disruptive signals that were not associated with the hemodynamic response by adding 5 confound dimensions.
5. Performed first level analysis (group ICA, PCA, and dynamic ICA analysis.)

 

Group PCA:

PCA is a dimensionality reduction technique that identifies the principal components (PCs) which represents the most significant variance across scans. The objective is to identify the commonalities and differences in brain connectivity of the patient group. It is exploratory and unsupervised, meaning it does not make prior assumptions about brain regions or networks. This allowed me to discover emergent patterns that are not biased by my hypotheses.

 

Group ICA:

ICA separates fMRI data into independent spatial components or networks, each representing a distinct pattern of brain activity. Like PCA, it is exploratory and was not based on a prior hypothesis.

 

Dynamic ICA:

Allowed me to compute temporally varying functional connectivity patterns by computing the differences in region-region connectivity over time. This allowed me to find variations in time series patterns between the groups, which can be revolutionary as it can demonstrate disruptions in connectivity that occur over time as well as which regions are disrupted during certain times.

 

These exploratory computations allowed me to adjust my hypothesis based on the results and identify regions of interest (ROIs). I observed significant disruptions particularly in regions implicated in GABAergic interneurons, such as the superior temporal gyrus and sensorimotor areas. Over time, the anterior and posterior STG showed unsynchronized activity which could mean there was a breakdown in normal brain network communication. Additionally, regions like the supramarginal gyrus and postcentral gyrus had time-dependent connectivity changes differing from the control group. This caused me to determine that GABAergic dysfunction may play a role in disrupted brain connectivity patterns in schizophrenia.

 

6. Created ROIs in FSLeyes based on GABAergic interneuron population as well as downloaded default ones from the Harvard-Oxford Cortical Atlas. 16 regions were determined to be of exceptional interest due to DOMINANT GABAergic populations. 
7. Determined criteria for accepting hypotheses and ensured ROIs included non cortical regions to ensure it is not just cortical disruption but GABAergic interneurons.
8. Performed Seed Based (SBC, specific brain regions that are pre-defined) and ROI-ROI based connectivity analysis.
9. Interpreted results in the context of GABAergic interneurons by consulting literature and deciding if results adhered to the criteria for accepting the hypothesis.

 

Single Cell RNA sequencing (scRNA seq) Analysis

 

Materials:

NCBI Gene Expression Omnibus

GALAXY Bioinformatics Software

DAVID Functional Annotation

NCBI Gene Database

 

1. scRNA raw reads of neural progenitor cells (2 control and 2 patient) were obtained from the NCBI Gene Expression Omnibus Bioproject PRJNA907299
2. Performed FASTQ to extract reads in GALAXY.
3. Performed quality assurance analysis (0 reads flagged for having <40 pHRED scores.
4. Aligned reads to human genome using HISAT2
5. Quantified gene expression using featureCounts
6. Quantified reads uploaded to DESEQ2
7. Organized upregulated and downregulated genes and discounted any with an adjusted p-value>0.05.
8. Performed pathway enrichment analysis on included genes using DAVID Functional Annotation and identified differentially expressed pathways.
9. Interpreted results using existing literature and NCBI Gene Database.

 

Designing the Prototype Pharmaceutical

Materials:

Marvin's CHEMdraw

SWISSADME

AutoDOCK Vina

PyMOL

NCBI Protein Data Bank

NovoPro Bioscience

OPSIN

Command Prompt on PC

 

After interpreting the results of multi-omic and fMRI analysis (more in research section as well as rationale for receptor targets), I determined that my pharmaceutical will focus on reducing neuroinflammation, regulating brain derived neuropathic factor pathways, activating 5-HT3, and partially agonising 5-HT1A.

This was achieved by starting with 7-8-Dihydroxyflavone and altering it in Marvin’s CHEMdraw software in the following manner:

 

A7: Deleted hydroxide

A8: Deleted hydroxide 

C3: Attached Fluorine

C4: Attached a Piperazine Ring

 

Converted new IUPAC name (3-fluoro-2-[3-methyl-4-(piperazin-2-yl)phenyl]-4H-chromen-4-one) into a SMILES code and conducted ADME analysis using SWISSADME to identify potential risks. No risk factors were identified so I continued to molecular docking.

1. Converted ligand (molecule created in CHEMdraw) into a pdbqt file in AutoDOCK Vina.
2. Deleted water, added polar hydrogens, and Kollman charges to receptors (TRKB, 5-HT3, 5-HT1A)
3. Created grid box with locations of the ligand and receptors.
4. Used command prompt to run binding affinity (kcal/mol).
5. Ran each test 3 times and used the average of runs when evaluating results.

I repeated these steps for three of the most common antipsychotics, Olanzapine, Clozapine, and Abilify.

Background

 

Introduction to schizophrenia

 

As mentioned in the problem segment, schizophrenia is a neurodevelopmental disorder affecting 24 million individuals worldwide and is characterized by psychotic symptoms such as hallucinations, delusions, and disorganized thinking as well as cognitive impairments which can impact an individual’s ability to integrate socially. With schizophrenia’s neurodevelopmental nature, there are numerous demographic differences, namely between male and female patients. The disorder typically emerges in early adulthood, with men experiencing a sharper peak in their early twenties, while women have a more gradual onset and a second, less pronounced peak in their midlife. The structural differences in the brains of schizophrenic patients are well documented, and most notably include ventricle enlargement, reductions in brain and grey matter volume, white matter integrity, and frontal and temporal lobe dysfunction. Neurochemically, schizophrenic patients are noted to experience dysregulation in dopamine signaling through hyperactivity in the mesolimbic pathway, as well as deficits in glutamatergic NMDA receptor function. 

 

Current treatment and limitations

 

(This section is inserted in the “problem” segment as well).

 

Currently, schizophrenia is being treated with antipsychotics which can be divided into first generation and second generation antipsychotics. As described previously, first generation antipsychotics targeted dopamine D2 receptors in the mesolimbic pathways which decreased positive symptoms but disregarded symptoms unrelated to excitatory cognitive function. However, these antipsychotics introduced extrapyramidal symptoms, disrupting nervous and cortical function causing parkinsons, tremors, and sensori-motor disruptions. Second generation antipsychotics, aim to ease symptoms by dually and partially antagonizing dopamine D2 receptors in the mesolimbic pathway and serotonin 5-HT2A receptors in the mesocortical pathway, which increases dopamine helping with social apathy as well as reduces tremors introduced by first generation medication [3, 4]. In some cases, benzodiazepines are prescribed as an adjunct treatment to manage agitation, catatonia, and anxiety. They enhance GABA-A receptor activity which increases neuronal inhibition and creates a calming effect [3, 4]. However, current medication is nonspecific and does not target the core etiology of schizophrenia, evident by the fact that one third of patients are treatment resistant to current antipsychotics [4]. Furthermore, the side effects are detrimental, making it difficult for patients to consistently stay on their medication. Side effects include neuromuscular dysfunction, skin and eye reactions, cardiovascular risks, sedation, increased risk of cognitive decline, and physical dependence [4]. One of the most overlooked symptoms, sexual dysfunction [4], is also one of the most common reasons for patients discontinuing treatment, which is vital for treatment success.

 

The most successful antipsychotics, including Clozapine and Olanzapine are ineffective over 30% of the time in patients. Treatment resistant individuals can also unfortunately be placed on long-term benzodiazepine treatment, which increases their risk for neurological dysfunction in adulthood as well as addiction and physical dependence.

 

Schizophrenia and Current Applications/Hypothesis' in fMRI

 

Neuroimaging is becoming an increasingly valuable tool in elucidating the etiology and neurodevelopment of schizophrenia as it is less invasive than other methods and also produces high resolution images, which allow researchers to work towards biomarker identification. Throughout the literature, high resolution MRI techniques have revealed that schizophrenia patient’s brain structures are significantly different from healthy controls but also differs during different stages of development. These changes include enlarged cerebrospinal fluid (CSF) spaces, reduced white and gray matter volumes, and specific volume losses in regions like the prefrontal cortex and thalamocortical connections. Moreover, these structural differences have guided the “disconnection theory” of schizophrenia which suggests that  histologically verified decreases in dendritic and synaptic density contribute to the social and cognitive symptoms of individuals with schizophrenia. The current hypothesis of development from a neuroimaging perspective involves the idea of abhorrent connectivity, meaning that functionally, schizophrenic patients are unable to synchronize cortical areas as well as overexcite certain regions unnecessarily. 

 

fMRI Analysis

 

fMRI is a non-invasive neuroimaging technique that can measure brain activity by quantifying blood flow over time. The brain is imaged in three-dimensional units called voxels, which are tiny cubes of tissue which become data points in an fMRI scan. Through rapid, successive measurements, fMRI captures time series data that reflect how brain activity fluctuates during both resting states and task-based experiments. This method allows scientists to map out functional networks and connectivity by examining how different regions (or “seeds”) correlate over time, with clusters of voxels showing synchronized activity suggesting functional partnerships. 

 

Hemodynamic Response

 

The Blood Oxygen Level-Dependent (BOLD) response is used to determine brain activity due to hemoglobins magnetic properties. When a brain region is activated, local neurons increase their firing rate, which requires a higher demand for oxygen and an increased cerebral metabolic rate of oxygen (CMRO₂). This new demand causes the arterial vessels to dilate and deliver more oxygenated blood than is needed. Oxygenated hemoglobin is diamagnetic, meaning it cannot be picked up as a signal, but deoxygenated hemoglobin is paramagnetic and alters the local magnetic environment which creates a quantifiable change in the MRI signal. This BOLD contrast can then be used to visualize which brain areas are active during specific tasks or even during rest.

 

Resting state analysis

 

Traditionally, fMRI analysis would be performed by getting participants to do a simple cognitive task to test a hypothesis regarding brain activity when engaged in certain activities. However, it was found that even during rest, there were discernible differences in BOLD signals across demographics. Resting state fMRI analysis (rsfMRI) captures the natural fluctuations in BOLD signals which can reveal how different brain regions inherently communicate which creates networks that are still notably active during rest.

 

This is incredibly beneficial in the context of schizophrenia, as it is characterized by disrupted neural networks and impaired communication between brain regions, and resting state allows us to observe these disruptions independent from a task.

 

fMRI Analysis Results

 

My project explored fMRI data of 72 controls and 75 individuals with schizophrenia obtained from the COBRE institute. Following exclusions detailed in my methodology, I processed the samples using the CONN toolbox and performed exploratory ICA analysis as well as seed and defined ROI-ROI analysis.

 

Exploratory results - Group ICA, PCA and Dynamic ICA

 

Group ICA and PCA Results

 

Independent Component Analysis (ICA) identified several components that align with known spatial brain networks. I was only able to investigate components that represented distinct functional networks in the brain, and because they adhered to already identified anatomical networks, I was able to deduce the functional disruptions in the schizophrenic group.

 

Below are the statistically significant networks identified, along with the brain regions associated with the component and my findings.

 

Identified Networks and Corresponding Components

Language Network (Component 7)

• 81% of the component covered 15% of the Middle Temporal Gyrus, posterior division (Left).
• 7% of the component covered 5% of the Superior Temporal Gyrus, posterior division (Left).

Aligns with known language-processing regions, particularly in the left hemisphere, which is critical for speech and comprehension.

Visual Network (Component 13)

• 50% covered 30% of the Intracalcarine Cortex (Right).
• 22% covered 15% of the Intracalcarine Cortex (Left).
• 6% covered 18% of the Supracalcarine Cortex (Right).
• 4% covered 4% of the Cuneal Cortex (Left).
• 2% covered 15% of the Supracalcarine Cortex (Left).
• 39% covered 22% of the Cuneal Cortex (Left).

This component primarily maps onto the visual cortex, particularly the intracalcarine and supracalcarine cortices, which play a role in early visual processing.

 Default Mode Network (Component 21)

• 32% covered 8% of the Insular Cortex (Right).
• 20% covered 8% of the Central Opercular Cortex (Right).
• 17% covered 15% of the Planum Polare (Right).
• 33% of voxels covered 18% of the Caudate (Right).
• 42% covered 19% of the Parietal Operculum Cortex (Right).
• 38% covered 21% of the Planum Temporale (Right).
• 65% covered 3% of the Lateral Occipital Cortex, superior division (Left).

The default mode network (DMN) is associated with self-referential thinking, memory, and mind-wandering. The identified regions suggest involvement in cognitive integration and resting-state activity.

Visual Network (Component 23)

• 58% of voxels covered 23% of the Thalamus (Left).
• 35% of voxels covered 15% of the Thalamus (Right).
• Decreased connectivity observed.

This component may be involved in visual processing and sensory relay. The decreased connectivity might indicate dysfunction in visual information processing or regulation.

Sensorimotor Network (Component 27)

• 54% of voxels covered 15% of the Supramarginal Gyrus, posterior division (Right).
• 23% of voxels covered 10% of the Supramarginal Gyrus, anterior division (Right).

The supramarginal gyrus is involved in sensorimotor integration and spatial awareness; this component plays a role in motor function and perception.

Sensorimotor Network (Component 37)

• 84% of voxels covered 11% of the Temporal Pole (Left).
• Other notable areas covered:
  • Parahippocampal Gyrus, anterior division (Left).
  • Temporal Fusiform Cortex, anterior division (Left).

This component highlights sensorimotor processing but also includes regions involved in higher-order cognitive and memory functions, such as the parahippocampal gyrus. 

• Multiple components correspond to well-known functional networks.
• Decreased connectivity in the thalamus (Component 23); possible impairments in sensory relay. 
• Sensorimotor networks (Components 27 & 37) have an involvement in temporal and parietal regions; movement coordination and spatial awareness.
• The default mode network shows strong involvement of insular, opercular, and occipital regions; role in cognitive and self-referential processing. 
• Individual regions were noted to have dominant GABAergic interneuron populations.

 

Dynamic ICA

 

Dynamic ICA revealed shifting connectivity patterns over time and reinforced the discovery from group ICA and PCA analysis as these patterns pertained specifically to regions rich in GABAergic interneurons, which are crucial for regulating excitatory-inhibitory balance in the brain. These findings further emphasize how disruptions in GABAergic signaling may contribute to schizophrenia-related dysconnectivity.

Statistically Significant Connectivity Changes Over Time

• Superior Temporal Gyrus (STG), Anterior Left: Showed a progressive decrease in connectivity with the lateral occipital cortex, visual lateral, and visual occipital regions. Notably, when one region was active, the other was not, so they had an asynchronous relationship rather than normal synchronization. Since both STG and these visual areas are GABAergic-rich, this disconnect highlights a potential failure in inhibitory regulation.
• Superior Temporal Gyrus, Posterior Left: In contrast, this region became increasingly connected to the lateral occipital cortex and occipital poles bilaterally, suggesting a compensatory mechanism or maladaptive reorganization in functional networks.
• Supramarginal Gyrus, Left: Decreasing connectivity over time with the lateral occipital cortex, inferior division, left, further supporting a pattern of disrupted functional integration.

Additional Connectivity Trends in Sensorimotor and Attention Networks

• Postcentral Gyrus, Left: Increasing connectivity with occipital fusiform gyrus, visual occipital, and occipital pole bilaterally.
• Precentral Gyrus, Left: Similarly, increased connectivity with occipital fusiform gyrus (both hemispheres), visual occipital, and occipital pole.
• Dorsal Attention Network (Frontal Eye Fields - FEF): Stronger connectivity with occipital pole left and fusiform gyrus left, highlighting changes in attention-related circuits.
• Juxtapositional Lobule Cortex, Left & Right: Both hemispheres showed increased connectivity to occipital regions, including the fusiform gyrus, lateral occipital cortex (inferior division), visual occipital, and occipital pole, reinforcing shifts in sensory integration mechanisms.

Like I deduced in group ICA and PCA analysis, GABAergic interneurons likely had a role in this time specific altered connectivity, and that cross-modal integration deficits may be influenced by impaired GABAergic modulation.

Additionally, increasing connectivity within the sensorimotor and attention networks may indicate compensatory neural mechanisms attempting to offset disruptions somewhere else. However, such reorganization is often maladaptive as it creates cognitive and perceptual disturbances.

 

Seed Based and ROI-ROI Connectivity

 

Through the exploratory methods, I developed a hypothesis centering around GABA-ergic interneuron disruption. This hypothesis is novel in that most research that implicates GABA aim to resolve the disruption by increasing the neurotransmitter itself. However, these results suggested a systemic failure in GABA-ergic rich areas, which pointed towards structure wide issues. The regions identified from ICA were particularly rich with GABA-ergic interneurons, and were also uniquely miscommunicating with eachother, often in compensetory ways overtime as revealed by dynamic ICA, revealing that this disruption is long standing in that patients have adjusted to the disruption with potentially damaging connectivity.

This allowed me to develop criteria that must be met to reject the null hypothesis when investigating seed and ROI-ROI connectivity:

1. Cortical areas must experience differing functional connectivity
2. Cortical ROIs pre-identified based on ICA results must show differing functional connectivity, themselves - other regions AND with each other
3. NON CORTICAL pre identified based on ICA results must show differing connectivity, WITH the pre identified gabaergic rich cortical ROIs.

16 regions of interest (ROIs) were identified based on their dominant GABAergic interneuron population. I selected neocortical regions as GABAergic interneurons are highly concentrated there, and particularly selected those originating from the medial and caudal ganglionic eminences. Moreover, I wanted to ensure disruptions were not just cortical, so I included hippocampal and parahippocampal areas involved in GABAergic regulation. Lastly, I excluded dopaminergically-dominant regions even if they had rich GABAergic populations.

 

Neocortical Regions 

• Primary Sensorimotor Areas:
  • Precentral Gyrus (PreCG)
  • Postcentral Gyrus (PostCG)
• Parietal and Multimodal Processing Regions:
  • Supramarginal Gyrus (SMG)
  • Superior Parietal Lobule (SPL)
• Visual Cortex:
  • Lateral Occipital Cortex (LOC)
  • Intracalcarine Cortex (ICC)
  • Occipital Fusiform Gyrus (OFusG)
  • Cuneal Cortex (Cuneal)
  • Occipital Pole (OP)
  • Supracalcarine Cortex

Non-Cortical Regions

Hippocampal and Parahippocampal Areas:

Hippocampus (Left & Right)

Anterior Parahippocampal Gyrus (aPaHC, Left & Right)

Posterior Parahippocampal Gyrus (pPaHC, Left & Right)

Higher-Order Integration Regions 

• Anterior Cingulate Cortex (ACC) 
• Anterior Insula (AInsula) 
• Insular Cortex (Left & Right)

 

Results

 

My analysis revealed that sensorimotor and visual cortical areas with a high density of GABAergic interneurons exhibited significantly greater disruptions in functional connectivity compared to non-GABAergic regions. These disruptions also extended beyond the cortex, as they also affected subcortical regions, particularly the hippocampus and striatum, both of which contain dominant GABAergic populations. Moreover, the abnormalities in these regions were more pronounced when examined in relation to my predefined regions of interest (ROIs).

Despite the widespread connectivity disruptions in GABAergic-implicated regions, no intra-ROI disruption was observed within the visual cortex itself. This means that specific gyri within the visual system did not show direct connectivity abnormalities among each other. However, the thalamus which is a critical relay structure crucial for GABAergic interneuron activity was implicated to have altered connectivity in all identified GABAergic rich regions. This suggests that while the visual cortex itself may not be directly impacted by GABAergic dysfunction, the systems it depends on are disrupted. These thalamic abnormalities likely contribute to impaired functional communication between the visual cortex and other brain areas.

 

Specific Region Findings

• Nearly all observed disruptions in thalamic connectivity were related to cortical regions, with 40% involving the 14 predefined GABAergic-rich ROIs.
• Altered connectivity patterns were observed with the thalamus and right posterior supramarginal gyrus; subcortical disruptions are not isolated in the cortex
  • One-third of the Precentral Gyrus’s altered connections involved GABAergic-rich ROIs.
  • Half of the Postcentral Gyrus’s disrupted connectivity patterns were linked to these regions.
  • The Posterior Supramarginal Gyrus had exclusive connectivity disruptions with highly GABAergic-implicated regions.
  • Superior Parietal Lobule & Lateral Occipital Cortex (Inferior Division): Both demonstrated significant alterations, with the thalamus consistently appearing as a region of disruption in their connectivity profiles.

 

Single Cell RNA Sequencing

 

Single cell RNA sequencing analysis (scRNA seq) is a sequencing method that investigates gene expression at the cellular level. While fMRI can measure functional connectivity between regions which allows assumptions to be made based on the functions of regions with altered connectivity, it cannot elucidate differences in cellular or even tissue wide function. Single cell RNA sequencing analysis can confirm the assertions drawn from fMRI analysis as if GABA-ergic dysfunction is truly present, genes and pathways vital for their success will likely experience downregulation. Additionally, it can supplement discoveries from fMRI analysis that were unclear or unable to be correlated with certainty, as well as highlight genome specific differences in expression when compared to healthy controls.

 

The Data

 

3 samples of neural progenitors obtained from the nasal canal of schizophrenic and healthy controls were downloaded from the NCBI Gene Expression Omnibus (GEO). As described in the methodology, RNA sequencing analysis as well as gene ontology analysis was performed.  A summary of the statistically significant results are provided below:

 

Downregulated Pathways/Processes

 

1. Axon Development
    
   • Essential for inhibitory neuron function; timing of circuit activity, inhibition type, and cortical computations.
   • Downregulation in axon development affects inhibitory GABAergic neurons.
2. Neurogenesis/Differentiation
    
   • Reduced expression of proteins involved in neural stem cell and neural progenitor cell differentiation.
   • Sp2 zinc-finger transcription factor downregulation disrupts the cell cycle, impairing neurogenesis.
   • Issues in differentiation of neurons may serve as an avenue for therapeutic intervention.

3. Retinoic Acid
 

• RA is essential for GABAergic neuron generation and is downregulated; incredibly important for neuroinflammation, oxidative stress, and neurodegeneration.
•  

4. Apoptosis
 

• Reduced apoptosis and disrupted morphogenesis affect neural patterning, migration, and differentiation.
• Misplaced GABAergic interneurons lead to improper inhibitory control and disorganized brain structures.

5. Energy Metabolism
 

• Downregulated pathways linked to cellular stress responses and metabolism indicate vulnerability to neurodevelopmental disruption.

 

Upregulated Pathways/Processes

 

Overgrowth, Proliferation:
 

• Unexpected upregulation of cell growth contradicts literature findings on reduced proliferation in schizophrenia.
• Suggests the presence of both overgrowth and dysregulated differentiation.

 

Neuroinflammation Immune Response:
 

• Increased expression of inflammation-related genes supports the neuroinflammatory hypothesis of schizophrenia.
• Elevated markers of immune activity correlate with abnormal neural development and function.

Signaling Pathways
 

Upregulation of MAPK signaling and other pathways linked to synaptic transmission, but unclear if compensatory or pathological.

 

I concluded through the analysis that my hypothesis stating that GABAergic interneurons were functionally disrupted rather than specifically up or downregulated is supported by omic analysis. Moreover, I concluded that neurogenesis and inflammation, particularly through retinoic acid pathways, were pertinent drug targets as GABAergic interneurons rely on neurodegeneration as well as steady immune responses, meaning that targeting those two could indirectly support GABAergic function.

 

Drug Design

 

As my analysis revealed severe GABAergic dysfunction, downregulated neurogenesis, and increased neuroinflammation causing disruption in axon development, it was evident that treatment must increase adult neurogenesis while also working to reduce inflammation, potentially through the retinoic acid pathways observed to be downregulated. Rodent studies consistently showed that current antipsychotics did not increase adult neurogenesis as well as adult stem cell proliferation, showing they are ineffective at treating the downregulation of neurogenesis present in schizophrenic patients. I created a chart summarizing what my drug aims to target that current antipsychotics do not:

Target Issue	Possible Mechanism	Potential Drug Classes
GABAergic interneuron development	Promote GABAergic differentiation via BDNF or Wnt signaling	BDNF mimetics, Wnt activators
Neurogenesis	Increase hippocampal neurogenesis	5-HT receptor agonists (similar to antidepressants)
Inflammation	Reduce cytokine impact on neural growth	IL-6 inhibitors, COX-2 inhibitors
Glutamate imbalance	Enhance NMDA function	Glycine transporter inhibitors, NMDA co-agonists

Fig 1: Conclusion based on fMRI and scRNA seq analysis and possible treatment options dependending on root mechanism.

 

This allowed me to conclude that I would like to construct a hybrid drug that acts as both a Brain Derived Neuropathic Factor (BDNF) agonist and a mild serotonin modulator, which will indirectly assist in preventing neuroinflammation as regulation of proliferation and neurogenesis can prevent immunal imbalances causing an inflammatory protective response.

 

Direct BDNF agonism is difficult as BDNF is a large protein. However, smaller molecules such as flavone structures are known to bind well to TRKB receptors. This allowed me to discover that 7,8-Dihydroxyflavone was the best molecule to base my drug off of, as it not only binded well to TRKB and thus agonised BDNF, it is blood brain barrier permeable, and also has anti-inflammatory properties and is known to increase retinoic acid metabolism efficiency. TrkB receptors are highly expressed on parvalbumin (PV+) interneurons, which are important for maintaining cortical inhibition, gamma oscillations for cognition, and balancing excitatory-inhibitory neurotransmission. This is achieved by increasing GAD67 expression, which boosts GABA synthesis,  the maturation of perisomatic inhibitory synapses, and regulating fast-spiking interneurons.
 

Moreover, targeting 5-HT (serotonin) receptors and partially agonising them would assist in neurogenesis as well as increase BDNF when taking a psychedelic approach by targeting 5-HT2A as it can enhance excitatory GABA release. This must be controlled by also partially agnosing 5-HT3 to allow for fast ionotropic activation with this onset of GABA release. This required modulations on the rings of the flavone.

 

Modifications on Flavone Base

Fig 2: Numbering scheme of the flavone core structure, with hydroxyl substitutions at C7 and C8. The A, B, and C rings are labeled following IUPAC conventions.

 

Fluorine added at C3

Piperazine ring added at C4

Hydroxyl groups deleted from A7 and A8

 

Rationales

Fluorine at C3

Fluorine was added to increase the lipophilicity of the molecule allowing better blood brain barrier permeation. Fluorine can also enhance the binding affinity to TrkB due to its high electronegativity which increases attraction, stabilizing interactions and ultimately activation.
 

Fluorine is more stable metabolically as well, especially by phase II enzymes, which helps with the metabolic stability of the drug. This helps the compound remain active for a longer period in the body, especially in the brain.
 

 Piperazine Ring at C4 

Piperazine was added as it is a heterocyclic amine which has flexible and positively charged nitrogen atoms which allow for strong hydrogen bonding with TRKB and electrostatic interactions. 

Piperazine is also commonly found in many neuroactive drugs that work in serotonergic, dopaminergic, and GABAergic systems for its lipophilicity and stability.
 

Deleting Hydroxyl Groups at A7 and A8

The objective was to reduce the polar surface area of the molecule, decreasing solubility and increasing lipophilicity which improves the molecule's ability to cross the BBB because fewer hydroxyl groups decrease hydrogen acceptors and donors.

Hydroxyl groups also increase metabolic degradation, so deleting them increases the half-life and bioavailability.

Fig 3: Final structural diagram of drug design (IUPAC name: 3-fluoro-2-[3-methyl-4-(piperazin-2-yl)phenyl]-4H-chromen-4-one)

 

Absorption, Distribution, Metabolism and Excretion (ADME) Analysis

 

ADME analysis was performed in SWISSADME to ensure the drug was viable. I was mainly looking for molecular weight to be under 500 g/mol, with under 350 being ideal, a consensus Log Po/w (lipophilicity value) between 1.2-3.0 (based on the criteria for blood brain barrier permeability outlined by Waterhouse, 2003), and proper pharmacokinetic properties, ensuring GI absorption, and inhibition of CYP2D6 AND CYP3A4 as it can assist in metabolism of the molecule.

 

Fig 4: ADME results displayed in SWISSADME, showcasing high synthetic bioavailability and drug-like properties.

 

Molecular Docking in AutoDock

 

In the molecular docking experiment, I ran the docking simulations for three different drug candidates, comparing their binding affinities to the target receptor. The goal was to predict how well each drug interacts with the receptors I aimed to agonise, TrkB, HT1A, and HT3A receptors) and how strongly they bind.

 

The docking software calculated the binding affinity for each drug in its best binding pose. This value indicates how strongly the drug interacts with the receptor. Lower (more negative) binding energies indicate stronger binding. The higher the negative score, the more thermodynamically favorable the interaction was.

Fig 5: AutoDOCK Vina Docking results after performing three trials and averaging them for each trial.

 

My drug performed significantly better in terms of binding affinity across all three targets. For TrkB, my drug showed a much stronger binding affinity compared to Clozapine, Abilify, and Olanzapine.

In terms of HT1A, my drug again outperformed the others, indicating a better binding affinity for HT1A. For HT3A, my drug showed a superior binding affinity as well.

When comparing the other drugs, I noticed that while Clozapine, Abilify, and Olanzapine performed relatively well in specific targets, they weren't all-rounders. For instance, Clozapine had a strong binding affinity at HT1A but was weaker at TrkB and HT3A. Abilify had decent binding at TrkB but underperformed at HT1A and HT3A, and Olanzapine showed weaker interactions across all three targets. This suggests that while some of these drugs may be stronger in one area, they don’t perform equally well across all targets, which means they may not be as versatile in therapeutic applications that require multi-receptor targeting.

Fig 1: Preprocessing and quality assurance in the CONN toolbox.

Fig 2: Addition of covariates using CONN toolbox.

Fig 3: Connectivity map visualized using CONN toolbox - blue representing decreased connectivity during time series, red representing increased and possibly abhorrent connectivity.

Fig 4: Connectome ring visualized using CONN toolbox; showcases inter-ROI connectivity based on pre-defined Harvard-Oxford parcellations.

Fig 5: Visualization of connectivity through connectome parcellations.

Fig 6: 5_HT1A receptor in AutoDock Vina.

Fig 7: 5_HT1A receptor in Pymol.

Fig 8: Best binding position of drug prototype for 5_HT1A receptor.

 

Fig 9: TKRB receptor in AutoDock Vina.

 

Fig 10: 5_HT3 and prototype drug interacting during docking analysis in AutoDock Vina.

Example of output of binding affinity scores using AutoDock Vina, with the numbers representing binding poses, 1 being the most optimal and 10 being the least:

mode |   affinity | dist from best mode

     | (kcal/mol) | rmsd l.b.| rmsd u.b.

-----+------------+----------+----------

   1        -10.1      0.000      0.000

   2         -9.6     17.644     19.810

   3         -9.6     29.088     31.590

   4         -9.5      4.450      7.611

   5         -9.5     29.271     31.813

   6         -9.4      1.822      3.005

   7         -9.2     15.110     17.677

   8         -9.1     19.672     21.353

   9         -9.1     26.828     28.767

Writing output ... done.

Conclusion

Schizophrenia remains a debilitating neurodevelopmental psychiatric disorder with limited effective treatment options as most antipsychotics fail to target disrupters of its root pathophysiology. Current antipsychotic drugs primarily modulate dopamine and serotonin pathways but ignore recent findings regarding neuroinflammation, implicated GABA  A dysfunction, and cerebral density. Given these limitations, finding an alternative, neurobiologically precise treatment is incredibly pertinent to increase treatment adherence for schizophrenic patients. 

In my project, I performed fMRI and scRNA seq analysis to develop a neurobiologically precise prototype molecule that targetted key players in schizophrenic etiology based on my results. Initial, exploratory fMRI analysis revealed multiple brain networks showing altered connectivity in individuals with schizophrenia. These included the Language Network, Visual Network, Default Mode Network, and Sensorimotor Networks, with the latter showing disruptions in motor and sensory integration regions such as the Supramarginal Gyrus and Precentral Gyrus. The analysis also highlighted decreased connectivity in the thalamus which was consistently implicated in various disrupted functional networks. Moreover, these changes evolved throughout time series, as notably in the Superior Temporal Gyrus and the visual networks the decreased connectivity varied and also converged specifically with GABAergic interneuron rich areas, showing these changes were  long-standing and potentially due to compensatory reorganization that may be maladaptive. As for the second level seed based analysis, I selected my ROIs based on their high density of GABAergic interneurons which essentially contained neocortical regions, as these areas have a significant concentration of GABAergic neurons originating from the medial and caudal ganglionic eminences. Additionally, I included hippocampal and parahippocampal areas, as they are involved in GABAergic regulation. Lastly, I excluded dopaminergic ally-dominant regions to ensure my conclusions were not convoluted by the traditional dopamine-centered hypothesis. 

My analysis revealed that sensorimotor and visual cortical areas with a high density of GABAergic interneurons exhibited significantly greater disruptions in functional connectivity compared to non-GABAergic regions. These disruptions also extended beyond the cortex, as they also affected subcortical regions, particularly the hippocampus and striatum, both of which contain dominant GABAergic populations. Moreover, the abnormalities in these regions were more pronounced when examined in relation to my predefined regions of interest (ROIs). Nearly all observed disruptions in thalamic connectivity were related to cortical regions, with 40% involving the 14 predefined GABAergic-rich ROIs. Moreover, altered connectivity patterns were observed with the thalamus and right posterior supramarginal gyrus; subcortical disruptions are not isolated in the cortex

The last consideration for designing my drug was scRNA sequencing analysis, where I observed downregulation in processes involved in GABAergic interneuron function, such as axon development, neurogenesis, and retinoic acid pathways, all of which are important for proper neuronal differentiation and GABAergic neuron generation. There was also a reduction in apoptosis and disruptions in energy metabolism, which affected neural patterning and cortical computations. In contrast, there was an upregulation in processes related to cell proliferation and overgrowth, which suggests a possible dysregulation in differentiation. This contrasted from the literature, which warrants further investigation. Lastly, neuroinflammation was consistently upregulated, which can also be noted by a downregulation of retinoic acid pathways as well. 

As our analysis revealed severe GABAergic dysfunction, downregulated neurogenesis, and increased neuroinflammation causing disruption in axon development, it was evident that treatment must increase adult neurogenesis while also working to reduce inflammation, potentially through the retinoic acid pathways observed to be downregulated. Rodent studies consistently showed that current antipsychotics did not increase adult neurogenesis as well as adult stem cell proliferation, showing they are ATLEAST ineffective at treating the downregulation of neurogenesis present in schizophrenic patients. I proposed a hybrid drug that acts as both a BDNF agonist and mild serotonin modulator to prevent neuroinflammation by regulating neurogenesis and proliferation. Using 7,8-Dihydroxyflavone as the base, this molecule binds well to TRKB receptors, crosses the blood-brain barrier, has anti-inflammatory effects, and enhances retinoic acid metabolism. I made a few modifications on the molecule with a fluorine atom at C3 to improve lipophilicity and stability, a piperazine ring at C4 to  strengthen interactions with TRKB, and deleting hydroxyl groups at A7 and A8 to reduce polarity in order to improve BBB penetration and bioavailability.

Following molecular docking, it was evident that while Clozapine, Abilify, and Olanzapine performed relatively well in specific targets, they weren't all-rounders. For instance, Clozapine had a strong binding affinity at HT1A but was weaker at TrkB and HT3A. Abilify had decent binding at TrkB but underperformed at HT1A and HT3A, and Olanzapine showed weaker interactions across all three targets. The docking scores for my proposed molecule far surpassed those of typical second generation antipsychotics, showing both that current treatment does not effectively target the underlying etiology concluded from this project but also that increasing GABAergic interneuron functionality can reduce neuroinflammation and promote adult neurogenesis as well. 

Implications and Future Directions

This project is incredibly pertinent as through fMRI and multi omic analysis, I was able to derive common factors affecting the pathophysiology of schizophrenia, with both methods confirming each other and supplementing independent conclusions. However, my project applies these findings practically by targeting the current gap in schizophrenia treatment, which is primarily pharmaceuticals targeting specific receptors rather than relying on serotonergic and dopaminergic modulation. The novelty of my prototype drug lies in its dual mechanism, as it functions as both a BDNF agonist and a mild serotonin modulator. This approach targets the underlying dysregulation in GABAergic interneurons, promotes adult neurogenesis, and reduces neuroinflammation instead of attempting to solely agonise neurotransmitter receptors, which implies that its effects are likely more long lasting rather than the non-specific, temporary symptom relief provided by current antipsychotics as they fail to target the neuroinflammatory and neurodegenerative processes found in schizophrenic patients. As this drug is still a prototype, however, and will require several phases of iteration to begin practical application, it provides a direction for antipsychotics that targets GABAergic interneuron functionality, as even new antipsychotics centered around GABA aim to increase its production, which as concluded from my study, is not the solution to abhorrent connectivity from overall dysfunction of interneurons. 

To advance this project and its practical applications, I aim to reduce its molecular weight, as while it is under 350 g/mol, most BBB permeating molecules are slightly smaller to ensure effective travel. This can be done by researching alternatives to the Piperazine ring, or by constructing a more specific version of the Piperazine ring instead of relying on pre-existing structures. Moreover, while lipophilicity is also acceptable, I still wish to increase it slightly to ensure permeability. Attempts to do so have resulted in issues with inhibition and thus potential toxicity based on preliminary ADME analysis, making this a difficult yet important direction. This implies I must gain a better understanding regarding pharmakokinetics in order to advance my design to ensure it remains safe. Lastly, after applying these changes, the most comprehensive and realistic tests of drug efficacy would be in vitro trials after successfully constructing the molecule, which is the end goal of my project. 

Apart from the drug, an interesting result from scRNA seq analysis is the upregulation in processes related to cell proliferation and overgrowth and therefore dysregulation in differentiation. As stated previously, this has yet to be discussed directly in the literature, so exploring the cellular production of individuals with schizophrenia through proteomics would allow me to gain a deeper understanding of this result.

 

A full list of citations used throughout the project can be found in my log book. Only sources utilized for the construction of this projectboard have been included in this segment.

[1] Patel, K. R., Cherian, J., Gohil, K., & Atkinson, D. (2014, September). Schizophrenia: Overview and treatment options. P & T : a peer-reviewed journal for formulary management. https://pmc.ncbi.nlm.nih.gov/articles/PMC4159061/ 

[2] Fatemi, S. H., & Folsom, T. D. (2009, May). The neurodevelopmental hypothesis of schizophrenia, revisited. Schizophrenia bulletin. https://pmc.ncbi.nlm.nih.gov/articles/PMC2669580/ 

[3] Chokhawala, K. (2023, February 26). Antipsychotic medications. StatPearls [Internet]. https://www.ncbi.nlm.nih.gov/books/NBK519503/

[4] Pandey, A., & Kalita, K. N. (2022a, August 30). Treatment-resistant schizophrenia: How far have we traveled?. Frontiers in psychiatry. https://pmc.ncbi.nlm.nih.gov/articles/PMC9468267/

Bao, Y., Ibram, G., Blaner, W. S., Quesenberry, C. P., Shen, L., McKeague, I. W., Schaefer, C. A., Susser, E. S., & Brown, A. S. (2012, May). Low maternal retinol as a risk factor for schizophrenia in adult offspring. Schizophrenia research. https://pmc.ncbi.nlm.nih.gov/articles/PMC3520602/#:~:text=We%20found%20that%20low%20maternal,from%20a%20large%20birth%20cohort.

Reay, W. R., & Cairns, M. J. (2019, October 30). The role of the retinoids in schizophrenia: Genomic and clinical perspectives. Nature News. https://www.nature.com/articles/s41380-019-0566-2

Kahn, R. S., Sommer, I. E., Murray, R. M., Meyer-Lindenberg, A., Weinberger, D. R., Cannon, T. D., O'Donovan, M., Correll, C. U., Kane, J. M., van Os, J., & Insel, T. R. (2015). Schizophrenia. Nature reviews. Disease primers, 1, 15067. https://doi.org/10.1038/nrdp.2015.67 

Dabiri, M., Dehghani Firouzabadi, F., Yang, K., Barker, P. B., Lee, R. R., & Yousem, D. M. (2022). Neuroimaging in schizophrenia: A review article. Frontiers in neuroscience, 16, 1042814. https://doi.org/10.3389/fnins.2022.1042814

Gur, R. E., & Gur, R. C. (2010). Functional magnetic resonance imaging in schizophrenia. Dialogues in clinical neuroscience, 12(3), 333–343. https://doi.org/10.31887/DCNS.2010.12.3/rgur

Hamina, A., Taipale, H., Lieslehto, J., Lähteenvuo, M., Tanskanen, A., Mittendorfer-Rutz, E., & Tiihonen, J. (2024). Comparative Effectiveness of Antipsychotics in Patients With Schizophrenia Spectrum Disorder. JAMA network open, 7(10), e2438358. https://doi.org/10.1001/jamanetworkopen.2024.38358

Efthimiou, O., Taipale, H., Radua, J., Schneider-Thoma, J., Pinzón-Espinosa, J., Ortuño, M., Vinkers, C. H., Mittendorfer-Rutz, E., Cardoner, N., Tanskanen, A., Fusar-Poli, P., Cipriani, A., Vieta, E., Leucht, S., Tiihonen, J., & Luykx, J. J. (2024). Efficacy and effectiveness of antipsychotics in schizophrenia: network meta-analyses combining evidence from randomised controlled trials and real-world data. The lancet. Psychiatry, 11(2), 102–111. https://doi.org/10.1016/S2215-0366(23)00366-8 

 

Chat GPT was used to clarify concepts at times, but nothing AI generated has been included in the write up itself as well as on my posterboard. 

Thank you to the researchers at the Centers of Biomedical Research Excellence (COBRE) for providing me access with the fMRI data utilized in the project and for their contribution to open data science.

Despite not mentoring me on this project directly, working at the Precision Neurodevelopmental Lab under PI Sarah MacEachern has been monumental in nurturing my interest in neurodevelopmental disorders and motivated me to explore alternative treatments for the psychiatric side of neurodevelopmental disorders. 

Thank you to all my friends for supporting me everyday!