A proof-of-concept study: investigating the impact of COMT genotype and proline on negative symptoms in Alzheimer’s disease Authors: James D. Clelland, Hector W. Cure, Ashley M. Canizares, Julia H. Anderson, Bimala Rawal, Sabrina A. Wong, Nancy Kerner, Edward D. Huey, Lawrence S. Honig, Devangere P. Devanand, Catherine L. Clelland Published: 2026 ABSTRACT BackgroundNegative neuropsychiatric symptoms, such as apathy, are a core feature of Alzheimer’s disease. Previous studies have shown that levels of fasting plasma proline and differential activity of the catechol-O-methyltransferase (COMT) enzyme, which metabolises dopamine, influence negative symptoms in patients with severe psychiatric illness and those at risk for psychosis. For patients with the COMT high activity enzyme (as assessed via the COMT Val158Met polymorphism), high plasma proline was associated with less severe negative symptoms. Conversely, high proline was associated with more severe negative symptoms in patients with the low activity COMT enzyme.AimsIn this proof-of-concept cross-sectional study, we tested the hypothesis that proline and COMT Val158Met interact to modify negative symptom severity across neuropsychiatric disease, specifically now investigating patients with Alzheimer’s disease dementia.MethodLeast Absolute Shrinkage and Selection Operator regression was employed to model the interaction between proline and COMT on negative symptoms in n = 50 patients with probable Alzheimer’s disease or mild cognitive impairment with underlying Alzheimer’s disease biomarkers.ResultsThe proline × COMT interaction significantly predicted symptoms as assessed via the negative items of the Positive and Negative Symptom Scale, interaction coefficient 0.025, p = 0.031, with a trend toward significance when assessed via the Scale for Assessment of Negative Symptoms in Alzheimer’s disease, interaction coefficient 0.075, p = 0.055. Higher proline was beneficial for both Val/Val and Val/Met dementia patients, but detrimental to patients with the low activity Met/Met COMT enzyme.ConclusionsHigher proline also has opposing effects on negative symptoms by COMT genotype in patients with dementia and further supports the development of therapeutics aimed at modulating this interaction pathway across neuropsychiatric disorders. Negative neuropsychiatric symptoms, such as apathy (characterised by the loss of motivation to participate in activities, social withdrawal and emotional indifference), blunted affect and alogia, are a core feature of Alzheimer’s disease, and constitute a behavioural dimension independent of depression and cognitive status. 1–3 Patients with mild cognitive impairment (MCI) also manifest negative neuropsychiatric symptoms. 4 Negative symptoms contribute substantially to the huge personal and economic costs of dementia: they are associated with functional deficits, 5 and a rapid course of both cognitive and functional decline. 6,7 Additionally, caregiver burden and distress has been significantly associated with the presence of negative neuropsychiatric symptoms in dementia patients. 8,9 Negative neuropsychiatric symptoms are also a hallmark of severe psychiatric disorders like schizophrenia. In our previous studies, we demonstrated that fasting peripheral levels of the amino acid proline, a central nervous system (CNS) neuromodulator, 10 interact with the activity of the catechol-O-methyltransferase (COMT) enzyme that inactives dopamine, to predict negative symptom outcomes in patients with, 11 or those at risk for, 12 psychiatric illnesses. Specifically, the COMT Val158Met functional polymorphism has a direct impact on dopamine levels in the frontal cortex, 13,14 and we observed that for patients with the high activity enzyme (COMT Val/Val), high plasma proline was associated with less severe negative symptoms or greater symptom improvement over time, with the opposite finding in those with one or two copies of the low-activity COMT Met allele, who exhibited significantly greater negative symptom severity in the presence of high proline. 11,12 Proline is catabolised by the enzyme proline oxidase, encoded by the proline dehydrogenase (PRODH) gene, and study of the Prodh null mouse suggests that elevated proline induces glutamatergic synaptic transmission, 15,16 and notably, results in increased prefrontal dopamine transmission and functional hyper-dopaminergic responses. A concomitant upregulation of COMT gene expression observed in the CNS, potentially reflects a compensatory response to the activated dopamine signalling in this model. 16 Elevated peripheral proline has been associated with multiple psychiatric phenotypes, 17,18 and there is evidence of increased CNS and peripheral proline levels in patients with Alzheimer’s disease. 19–21 Furthermore, while a reduction in dopamine, dopamine metabolites and dopamine receptors have been reported in Alzheimer’s disease (reviewed in 22 ), studies have described associations between COMT and psychosis in Alzheimer’s patients 23 (and reviewed in 24 ), and the Val158Met genotype has been associated with neuropsychiatric symptoms and degeneration within dopamine-innervated brain structures (such as the prefrontal cortex) in patients with multiple dementias. 25 Given these findings, we tested the hypothesis that dopamine (as assessed via COMT Val158Met) and proline interact to also modify negative symptom severity in patients with Alzheimer’s disease. Method We conducted a cross-sectional study of patients with a clinical diagnosis of probable Alzheimer’s disease or mild cognitive impairment with underlying Alzheimer’s biomarkers (MCI+). All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2013. The protocol was approved by the Columbia University Institutional Review Board (IRB) (numbers AAAR8755, AAAS0071 and AAAV5130), and by the US Army Medical Research and Development Command, Office of Human and Animal Research Oversight. Eligible subjects were identified from participants of Columbia University’s Alzheimer’s and Dementia Research Center (ADRC), Columbia’s Memory Disorders Clinic and from patients under the care of the Columbia Doctors Neurology Aging and Dementia Practice. All subjects were interviewed as to their capacity to provide informed consent (see the supplementary material available at https://doi.org/10.1192/bjo.2025.10926 for additional details), and written informed consent was obtained from all subjects. Subjects were instructed to fast 8+ hours prior to their on-site single-study visit. Negative symptoms were assessed via the Positive and Negative Symptom Scale (PANSS) and the Scale for the Assessment of Negative Symptoms in AD (SANS-AD). 1 Positive PANSS symptoms were also evaluated. The Geriatric Depression Scale (GDS) and Mini-Mental State Examination (MMSE) were employed to assess depression and cognition, respectively (see the supplementary material for details on assessment administration). A fasting blood draw was obtained for assay of plasma proline and COMT genotyping, as described. 11,12 Demographics, clinical characteristics and medication status were compared across genotypes, and genotype distributions tested for Hardy–Weinberg equilibrium (HWE). The relationship between negative symptoms, demographics and clinical characteristics, was also assessed. Testing the primary hypothesis of an interaction between COMT and proline on negative symptoms and following the statistical analysis design of our previous study, 12 we employed Least Absolute Shrinkage and Selection Operator (Lasso) regression for model prediction, using the double selection feature for testing a priori determined covariates (gender, depression and positive symptoms). Again, as a sensitivity analysis, we employed a post-hoc stepwise selection procedure, starting with the independent variables of proline, COMT and the COMT × proline interaction, plus Lasso retained covariates. Analysis was conducted in Stata/BE v18.0 for Windows (College Station, Texas, USA; https://www.stata.com/). For all tests, the level of significance was fixed at p < 0.05, two-tailed. Benjamini–Hochberg’s corrections were employed to control the False Discovery Rate (FDR, 5%) following use of three negative symptom assessment scales. STROBE reporting guidelines were employed. Results A total of 50 participants completed the study. Demographic and clinical characteristics are shown in Table 1. The sample was in HWE for Val158Met (p > 0.05), and genotype groups were well matched. Regarding negative symptoms, there were no differences across genotypes for the negative items of the PANSS (sum of items N1, N2, G7, G8 and G10), 26 or the PANSS negative subscale (sum of items N1–N7). However, Val/Val subjects had significantly higher total SANS-AD scores (Table 1, p = 0.044) and so as for our previous study of psychiatric patients, 11 we employed the negative items of the PANSS as the primary outcome assessment, with secondary analyses of negative symptoms using the SANS-AD and the PANSS negative subscale. For all negative symptom assessments, significantly higher scores were observed in Alzheimer’s disease as compared with MCI + dementia (Table 1), and therefore diagnosis was included as a covariate. Further, and as expected, subjects with Alzheimer’s disease had significantly lower MMSE scores compared with subjects with MCI + dementia (Supplementary Table S1, p = 0.0006), and to avoid multicollinearity between the independent but related variables of diagnosis and cognition, diagnosis was chosen as the covariate for all subsequent models. Table 1Demographic and clinical characteristics of the dementia subjects, n = 50CharacteristicVal/Val (n = 18)Val/Met (n = 24)Met/Met (n = 8)Probability a Gender, n (female/male)10/816/86/20.63Race, n 0.11 African American420 Caucasian12227 Other/Not reported201Ethnicity, n (Hispanic or Latino, Y/N)7/118/162/60.85Age b (years), mean ± s.d.72.2 ± 7.672.0 ± 8.066.1 ± 7.10.18Education c (years), mean ± s.d.13.6 ± 4.415.4 ± 2.615.9 ± 3.30.19Testing language (English/Spanish)14/418/67/10.90Diagnosis, n (Alzheimer’s/MCI+)15/319/57/11.00Fasting plasma proline (μMol/L), mean ± s.d.203.3 ± 53.0228.1 ± 65.2220.5 ± 85.60.47Symptoms Negative items d (PANSS), mean ± s.d.15.0 ± 5.812.1 ± 5.012.9 ± 3.70.20 SANS-AD total, mean ± s.d.37.1 ± 18.425.8 ± 18.019.8 ± 15.2 0.044* PANSS Negative Subscale, mean ± s.d.21.5 ± 8.618.5 ± 8.119.0 ± 5.70.46 PANSS Positive Subscale, mean ± s.d.11.3 ± 4.210.9 ± 4.111.3 ± 4.30.96 GDS total, mean ± s.d.3.6 ± 3.12.7 ± 3.02.5 ± 3.60.55 MMSE total, mean ± s.d.18.8 ± 6.619.8 ± 5.717.4 ± 4.20.60Medications e Antidepressant, n (Y/N)6/117/173/50.85 Antihypertensive, n (Y/N)9/88/164/40.40 Antilipidemic, n (Y/N)9/89/151/70.16 Cognitive enhancer, n (Y/N)13/416/85/30.77 Neuroleptic, n (Y/N)4/132/220/80.23 Supplement, n (Y/N)10/714/104/40.93MCI, mild cognitive impairment; PANSS, Positive and Negative Symptom Scale; SANS-AD, Scale for the Assessment of Negative Symptoms in Alzheimer’s Disease; GDS, Geriatric Depression Scale; MMSE, Mini-Mental State Examination.Bold text indicates a significant difference between groups.a.Significant p-value (<0.05) when comparing characteristics across three catechol-O-methyltransferase genotype groups, calculated by one-way analysis of variance, Kruskal–Wallis or Fisher’s exact test, as appropriate. All tests were two-tailed.b.n = 49, as one subject did not report their age.c.n = 48, as two subjects did not report their years of education.d.Items N1, N2, G7, G8 and G10 of the PANSS.e. n = 49, as medication use for one subject was not available. Only medication groups prescribed to ≥10% of the sample (n = 5) are reported. Table 1Demographic and clinical characteristics of the dementia subjects, n = 50CharacteristicVal/Val (n = 18)Val/Met (n = 24)Met/Met (n = 8)Probability a Gender, n (female/male)10/816/86/20.63Race, n 0.11 African American420 Caucasian12227 Other/Not reported201Ethnicity, n (Hispanic or Latino, Y/N)7/118/162/60.85Age b (years), mean ± s.d.72.2 ± 7.672.0 ± 8.066.1 ± 7.10.18Education c (years), mean ± s.d.13.6 ± 4.415.4 ± 2.615.9 ± 3.30.19Testing language (English/Spanish)14/418/67/10.90Diagnosis, n (Alzheimer’s/MCI+)15/319/57/11.00Fasting plasma proline (μMol/L), mean ± s.d.203.3 ± 53.0228.1 ± 65.2220.5 ± 85.60.47Symptoms Negative items d (PANSS), mean ± s.d.15.0 ± 5.812.1 ± 5.012.9 ± 3.70.20 SANS-AD total, mean ± s.d.37.1 ± 18.425.8 ± 18.019.8 ± 15.2 0.044* PANSS Negative Subscale, mean ± s.d.21.5 ± 8.618.5 ± 8.119.0 ± 5.70.46 PANSS Positive Subscale, mean ± s.d.11.3 ± 4.210.9 ± 4.111.3 ± 4.30.96 GDS total, mean ± s.d.3.6 ± 3.12.7 ± 3.02.5 ± 3.60.55 MMSE total, mean ± s.d.18.8 ± 6.619.8 ± 5.717.4 ± 4.20.60Medications e Antidepressant, n (Y/N)6/117/173/50.85 Antihypertensive, n (Y/N)9/88/164/40.40 Antilipidemic, n (Y/N)9/89/151/70.16 Cognitive enhancer, n (Y/N)13/416/85/30.77 Neuroleptic, n (Y/N)4/132/220/80.23 Supplement, n (Y/N)10/714/104/40.93MCI, mild cognitive impairment; PANSS, Positive and Negative Symptom Scale; SANS-AD, Scale for the Assessment of Negative Symptoms in Alzheimer’s Disease; GDS, Geriatric Depression Scale; MMSE, Mini-Mental State Examination.Bold text indicates a significant difference between groups.a.Significant p-value (<0.05) when comparing characteristics across three catechol-O-methyltransferase genotype groups, calculated by one-way analysis of variance, Kruskal–Wallis or Fisher’s exact test, as appropriate. All tests were two-tailed.b.n = 49, as one subject did not report their age.c.n = 48, as two subjects did not report their years of education.d.Items N1, N2, G7, G8 and G10 of the PANSS.e. n = 49, as medication use for one subject was not available. Only medication groups prescribed to ≥10% of the sample (n = 5) are reported. Demographic and clinical characteristics of the dementia subjects, n = 50 MCI, mild cognitive impairment; PANSS, Positive and Negative Symptom Scale; SANS-AD, Scale for the Assessment of Negative Symptoms in Alzheimer’s Disease; GDS, Geriatric Depression Scale; MMSE, Mini-Mental State Examination. Bold text indicates a significant difference between groups. Significant p-value (<0.05) when comparing characteristics across three catechol-O-methyltransferase genotype groups, calculated by one-way analysis of variance, Kruskal–Wallis or Fisher’s exact test, as appropriate. All tests were two-tailed. n = 49, as one subject did not report their age. n = 48, as two subjects did not report their years of education. Items N1, N2, G7, G8 and G10 of the PANSS. n = 49, as medication use for one subject was not available. Only medication groups prescribed to ≥10% of the sample (n = 5) are reported. Via Lasso regression modelling, we observed a significant interaction between COMT and fasting plasma proline on negative symptoms (as assessed via the negative items of the PANSS 26 ), following adjustment for diagnosis and the retained variable of the PANSS positive subscale (Table 1, interaction β-coefficient 0.025, 95% CI (0.002, 0.047), p = 0.031 which remained significant after testing correction). Higher proline was beneficial for both Val/Val and Val/Met dementia patients, but detrimental to Met/Met patients (Fig. 1). In a post-hoc sensitivity analysis the interaction remained significant (Supplementary Table S2), suggesting a robust finding. In secondary analyses, we tested for an interaction between COMT and proline on two additional and highly related measures of negative symptoms, the SANS-AD total score and the PANSS negative subscale; however, although supportive of our primary analysis, the interactions did not reach statistical significance (SANS-AD interaction β-coefficient 0.075, 95% CI (−0.002, 0.153), p = 0.055, Table 2 and Supplementary Fig. S1; PANSS negative subscale interaction β-coefficient 0.030, 95% CI (−0.005, 0.066), p = 0.093). Fig. 1The interaction between catechol-O-methyltransferase genotype and proline on the negative items of the Positive and Negative Symptom Scale (PANSS) subscale. Data are plotted for each genotype group: Met/Met (left panel, n = 8, red), Val/Val (middle panel, n = 18, green), and Val/Met (right panel, n = 24, blue). Regression lines represent predicted values from simple linear models, with shaded areas indicating 95% confidence intervals. Among individuals with the Met/Met genotype, higher proline levels (x-axis) are associated with more severe negative symptoms (y-axis). In contrast, an inverse relationship is observed in both the Val/Val and Val/Met groups, where elevated proline levels correspond to fewer or less severe negative symptoms. The interaction between catechol-O-methyltransferase genotype and proline on the negative items of the Positive and Negative Symptom Scale (PANSS) subscale. Data are plotted for each genotype group: Met/Met (left panel, n = 8, red), Val/Val (middle panel, n = 18, green), and Val/Met (right panel, n = 24, blue). Regression lines represent predicted values from simple linear models, with shaded areas indicating 95% confidence intervals. Among individuals with the Met/Met genotype, higher proline levels (x-axis) are associated with more severe negative symptoms (y-axis). In contrast, an inverse relationship is observed in both the Val/Val and Val/Met groups, where elevated proline levels correspond to fewer or less severe negative symptoms. Table 2Prediction of negative symptoms by the COMT × proline interaction, n = 50Dependent/retained independent variables β (95% CI)s.e. z p-valueModel 1: Negative items (PANSS) a Proline−0.054 (−0.102, −0.007)0.024−2.240.025 COMT−6.510 (−12.06, −0.957)2.833−2.300.022 COMT × proline0.025 (0.002, 0.047)0.0112.160.031 Diagnosis−2.960 (−6.053, 0.133)1.578−1.880.061 PANSS positive subscalen/rModel 2: SANS-AD total b Proline−0.165 (−0.318, −0.011)0.078−2.110.035 COMT−24.220 (−44.24, −4.202)10.21−2.370.018 COMT × proline0.075 (−0.002, 0.153)0.0391.920.055 Diagnosis−12.12 (−23.10, −1.135)5.602−2.160.031 GDS total1.089 (−0.620, 2.798)0.8721.250.212 PANSS positive subscalen/rCOMT, catechol-O-methyltransferase; s.e. of β; z, critical value; PANSS, Positive and Negative Symptom Scale; n/r, point estimates of retained covariate(s) not reported in Lasso model; SANS-AD, Scale for the Assessment of Negative Symptoms in AD; GDS, Geriatric Depression Scale.a.Model 1, Wald χ 2(4) = 10.14, p = 0.0382. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary). Covariates: GDS total score (continuous), gender (binary), PANSS positive subscale (continuous).b.Model 2, Wald χ 2(5) = 23.94, p = 0.0002. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary), GDS total score (continuous). Covariates: gender (binary), PANSS positive subscale (continuous). Table 2Prediction of negative symptoms by the COMT × proline interaction, n = 50Dependent/retained independent variables β (95% CI)s.e. z p-valueModel 1: Negative items (PANSS) a Proline−0.054 (−0.102, −0.007)0.024−2.240.025 COMT−6.510 (−12.06, −0.957)2.833−2.300.022 COMT × proline0.025 (0.002, 0.047)0.0112.160.031 Diagnosis−2.960 (−6.053, 0.133)1.578−1.880.061 PANSS positive subscalen/rModel 2: SANS-AD total b Proline−0.165 (−0.318, −0.011)0.078−2.110.035 COMT−24.220 (−44.24, −4.202)10.21−2.370.018 COMT × proline0.075 (−0.002, 0.153)0.0391.920.055 Diagnosis−12.12 (−23.10, −1.135)5.602−2.160.031 GDS total1.089 (−0.620, 2.798)0.8721.250.212 PANSS positive subscalen/rCOMT, catechol-O-methyltransferase; s.e. of β; z, critical value; PANSS, Positive and Negative Symptom Scale; n/r, point estimates of retained covariate(s) not reported in Lasso model; SANS-AD, Scale for the Assessment of Negative Symptoms in AD; GDS, Geriatric Depression Scale.a.Model 1, Wald χ 2(4) = 10.14, p = 0.0382. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary). Covariates: GDS total score (continuous), gender (binary), PANSS positive subscale (continuous).b.Model 2, Wald χ 2(5) = 23.94, p = 0.0002. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary), GDS total score (continuous). Covariates: gender (binary), PANSS positive subscale (continuous). Prediction of negative symptoms by the COMT × proline interaction, n = 50 COMT, catechol-O-methyltransferase; s.e. of β; z, critical value; PANSS, Positive and Negative Symptom Scale; n/r, point estimates of retained covariate(s) not reported in Lasso model; SANS-AD, Scale for the Assessment of Negative Symptoms in AD; GDS, Geriatric Depression Scale. Model 1, Wald χ 2(4) = 10.14, p = 0.0382. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary). Covariates: GDS total score (continuous), gender (binary), PANSS positive subscale (continuous). Model 2, Wald χ 2(5) = 23.94, p = 0.0002. Independent variables: COMT (ordinal Val/Val, Val/Met, Met/Met), proline (continuous), interaction (COMT × proline), diagnosis (binary), GDS total score (continuous). Covariates: gender (binary), PANSS positive subscale (continuous). Discussion In this proof-of-concept study of patients with Alzheimer’s disease or MCI dementia with underlying Alzheimer’s disease biomarkers, we observed that levels of the CNS neuromodulator proline and COMT Val158Met genotype interact to predict negative neuropsychiatric symptoms. As for our studies of patients with severe psychiatric illness 11 and those at risk for psychosis, 12 higher proline was beneficial for Val/Val dementia patients, but detrimental to those with the Met/Met genotype. Interestingly, in this first study of dementia patients, high proline was also found to be beneficial for Val/Met patients, which is somewhat inconsistent with our prior findings on psychiatric patients. 11,12 Elevated plasma proline (hyperprolinemia) has been linked to a range of neuropsychiatric phenotypes, including developmental delay and intellectual disability, autism spectrum disorder, psychosis spectrum disorders 17,18 and Alzheimer’s disease. 19–21 Further, in patients with 22q11 deletion syndrome, who routinely exhibit elevated proline due to hemizygous deletion of the PRODH gene on chromosome 22, negative neuropsychiatric symptoms show opposing severities by COMT allele. 27–31 Taken together, we suggest a common mechanism across neuropsychiatric disorders, as previously proposed 11,12 : specifically, for Met/Met patients in the presence of high proline, a potential increase in dopamine transmission in the prefrontal cortex is exacerbated by low COMT activity, ultimately resulting in a frontal dopaminergic state above optimal levels. Conversely, in Val/Val patients, higher prefrontal COMT activity would likely reduce dopamine, limiting dopamine-receptor-mediated excitation, 32,33 and leading to potential prefrontal hypodopaminergia. Thus, higher proline would be beneficial for Val/Val patients, via shifting prefrontal dopamine signalling to optimal levels. In this model, and as speculated, 11,12,34 negative symptoms would be induced in conditions of both hyper- and hypo-dopaminergia, reflecting the inverted U-shape curve proposed for cognitive dysfunction in psychosis, 33 executive function in Parkinson’s disease 35 and the opposite impact of dopamine receptor availability on cognitive and motor performance in Alzheimer’s disease. 36 It is interesting to speculate that reduced catecholamine levels in the ageing cortex, 37,38 may dampen the impact of elevated proline, particularly for those patients with the Val/Met genotype who have intermediate enzyme activity, resulting in the unexpected finding of a negative relationship between high proline and less severe symptoms in heterozygotes. Regarding limitations, one key limitation of this study is the sample size; as completion of 50 participants in total resulted in relatively small genotype groups. However, this first investigation of dementia patients builds positively from our previous studies. Although the COMT × proline interaction did not reach statistical significance for all negative symptom assessments, we did observe a significant interaction using the negative items of the PANSS, the one assessment that was used across all our studies, 11,12 with interactions trending towards significance when measuring negative symptoms via the SANS-AD total score and the PANSS negative subscale. The SANS-AD was specifically designed for research with dementia patients; the scale derived from the original Scale for Assessment of Negative Symptoms, but with the attention and alogia sections removed. 1 The difference in effect sizes and significance between the negative items of the PANSS, the PANSS negative subscale and the SANS-AD may indeed reflect the different behaviours assessed. Replication of this work with a larger dementia sample would thus be valuable to fully explore these initial findings, in particular using instruments specifically developed for individuals with neurodegenerative diseases, 39 while also accounting for cultural factors. 40 Future work should also include mechanistic studies to validate the proline x dopamine interaction hypothesis. In summary, our exploratory, but positive finding that high proline has opposing effects on negative symptoms by COMT genotype in patients with dementia, further supports the development of therapeutics to specifically target the interaction pathway, across neuropsychiatric disorders. Supporting information Clelland et al. supplementary material Clelland et al. supplementary material