Emergence of PGx in psychiatry

The events leading to the emergence of PGx in psychiatry began in ancient Greece and has been described in detail elsewhere (Ampong, Reference Ampong2019; Müller & Rizhanovsky, Reference Müller and Rizhanovsky2020). In brief, the Greek scientist Pythagoras described an adverse reaction, now known as acute hemolytic anemia, that was experienced by some, but not all, people who consumed fava beans. Although at that time Pythagoras was unaware that genetic variants in the gene encoding glucose-6-phosphate dehydrogenase (G6PD) caused this adverse reaction, it is often considered the first PGx observation (Meletis, Reference Meletis2012). However, the foundation for modern PGx was not established until the twentieth century, when the seminal works of Sir Archibald Garrod (Prasad & Galbraith, Reference Prasad and Galbraith2005), Arno Motulsky (Motulsky, Reference Motulsky1957), Friedrich Vogel (Vogel, Reference Vogel1959), and Werner Kalow (Kalow, Reference Kalow1962) were published. These works coupled with the completion of the Human Genome project in 2003 facilitated a series of key events that led to the current application of PGx in psychiatry (Fig. 1).

Figure 1. Major events in the emergence of pharmacogenomics in psychiatry. AMP, Association of Molecular Pathology; CPIC, Clinical Pharmacogenetics Implementation Consortium; CPNDS, Canadian Pharmacogenomics Network for Drug Safety; DPWG, Dutch Pharmacogenetics Working Group; FDA, Food and Drug Administration; ISPG, International Society of Psychiatric Genetics; PGRN, Pharmacogenomics Global Research Network; PGx, pharmacogenomic; PharmGKB, Pharmacogenomics Knowledge Base; PharmVar, Pharmacogene Variation Consortium.

Among the most influential of these key events was the establishment of PGx implementation consortia, such as the Clinical Pharmacogenetics Implementation Consortium (CPIC) and the Dutch Pharmacogenetics Working Group (DPWG). These consortia were formed to facilitate the use of PGx testing for patient care via development of evidence-based, peer-reviewed, and freely available clinical prescribing guidelines. The first PGx-based dosing guidelines relevant to psychiatry were published in 2008 by the DPWG and included dosing recommendations for antidepressants (i.e. clomipramine, imipramine, nortriptyline, paroxetine, and venlafaxine), antipsychotics (i.e. haloperidol, risperidone, and zuclopenthixol), and one medication (i.e. atomoxetine) for attention-deficit hyperactivity disorder (ADHD) based on CYP2D6 genotype (Swen et al., Reference Swen, Wilting, de Goede, Grandia, Mulder, Touw and Deneer2008). In 2011, these guidelines were updated, and dosing recommendations were included for additional antidepressants (i.e. amitriptyline, citalopram, doxepin, escitalopram, and sertraline) and antipsychotics (i.e. aripiprazole, fluphenthixol) based on CYP2C19 or CYP2D6 genotype (Swen et al., Reference Swen, Nijenhuis, de Boer, Grandia, Maitland-van der Zee, Mulder and Guchelaar2011). CPIC followed with dosing guidelines of their own for tricyclic antidepressants (Hicks et al., Reference Hicks, Swen, Thorn, Sangkuhl, Kharasch and Ellingrod2013) and carbamazepine (Leckband et al., Reference Leckband, Kelsoe, Dunnenberger, George, Tran and Berger2013) in 2013 as well as a guideline for selective serotonin reuptake inhibitors in 2015 (Hicks et al., Reference Hicks, Bishop, Sangkuhl, Müller, Ji and Leckband2015) and atomoxetine in 2019 (Brown et al., Reference Brown, Bishop, Sangkuhl, Nurmi, Mueller, Dinh and Leeder2019). Updates to the CPIC tricyclic antidepressant, carbamazepine/oxcarbazepine, and serotonin reuptake inhibitor antidepressant guidelines were published in 2016 (Hicks et al., Reference Hicks, Sangkuhl, Swen, Ellingrod, Müller, Shimoda and Stingl2016), 2018 (Phillips et al., Reference Phillips, Sukasem, Whirl-Carrillo, Müller, Dunnenberger, Chantratita and Pirmohamed2018), and 2023 (Bousman et al., Reference Bousman, Stevenson, Ramsey, Sangkuhl, Hicks, Strawn and Bishop2023), respectively. In 2023, the DPWG again updated their antipsychotic prescribing guidelines, adding additional dosing recommendations for brexpiprazole and pimozide based on CYP2D6 genotype and quetiapine based on CYP3A4 genotype (Beunk et al., Reference Beunk, Nijenhuis, Soree, de Boer-Veger, Buunk, Guchelaar and van der Weide2023). A CPIC antipsychotic dosing guideline is anticipated in 2024.

To date, CPIC and DPWG have collectively developed PGx-based dosing guidelines for 14 antidepressants, seven antipsychotics, three anticonvulsant/mood stabilizers, and one medication for ADHD (Table 1). Importantly, these guidelines have been endorsed by the American Society of Health-System Pharmacists (ASHP, n.d.), American Society for Clinical Pharmacology and Therapeutics (ASCPT, n.d.), Canadian Paediatric Society (Canadian Paediatric Society, 2023), as well as six European healthcare organizations including the European Association for Clinical Pharmacology and Therapeutics and the European Association of Hospital Pharmacists (‘Statement of support for the use of European pharmacogenomic guidelines,’, n.d.). Moreover, in 2020, the US Food and Drug Administration (FDA) issued clearance for 23andMe to offer CYP2C19 testing for citalopram without the need for confirmatory testing by an additional certified clinical laboratory, marking the first FDA clearance of a PGx test for a psychotropic medication (Food & Drug Administration, 2020). Finally, an expert group convened by the International Society of Psychiatric Genetics came to a consensus that the current published evidence supports the use of PGx testing for CYP2D6 and CYP2C19 to inform dosing of antidepressants, antipsychotics, and atomoxetine, as well as testing for HLA-A and HLA-B prior to initiation of carbamazepine or oxcarbazepine (Bousman et al., Reference Bousman, Bengesser, Aitchison, Amare, Aschauer, Baune and Müller2021).

Table 1. Drug–gene pairs with pharmacogenomic prescribing guidelines

IM, intermediate metabolizer; NM, normal metabolizer; PM, poor metabolizer; RM, rapid metabolizer; UM, ultrarapid metabolizer.

^a Recommendations are current as of 1 June 2023. Full recommendation, evidence reviews, and updates can be accessed via the Pharmacogenomics Knowledge Base: https://www.pharmgkb.org/

^b Information on assignment of level of evidence can be found here: https://www.pharmgkb.org/page/clinAnnLevels

Implementation of PGx testing in psychiatry

Despite the availability of PGx-based dosing guidelines, inclusion of PGx information in certain FDA-approved product labels, and endorsements from professional societies, the adoption of PGx testing into routine mental health care has been modest. On the Innovation Adoption Curve (Rogers, Reference Rogers and Orthner1981), PGx in psychiatry is in the ‘early adopter’ phase and approaching the ‘chasm’, where innovations either fail or succeed in gaining the traction required to cross into the ‘early majority’ phase and achieve sustainable adoption. Several hurdles to crossing the chasm have been identified and reviewed in detail elsewhere, including (1) uncertainty about the clinical efficacy and cost-effectiveness (Maruf & Bousman, Reference Maruf and Bousman2022; Morris et al., Reference Morris, Alsaidi, Verbyla, Cruz, Macfarlane, Bauer and Patel2022; Murphy, Fonseka, Bousman, & Müller, Reference Murphy, Fonseka, Bousman and Müller2022), (2) low perceived clinical generalizability (Bousman et al., Reference Bousman, Bengesser, Aitchison, Amare, Aschauer, Baune and Müller2021), (3) minimal standardization and regulation (Bousman & Dunlop, Reference Bousman and Dunlop2018; Bousman, Jaksa, & Pantelis, Reference Bousman, Jaksa and Pantelis2017; Fan & Bousman, Reference Fan and Bousman2019), (4) lack of incorporation into clinical practice guidelines (Maruf & Bousman, Reference Maruf and Bousman2022), (5) challenges with integration into electronic health/medical records (Caraballo et al., Reference Caraballo, Sutton, Giri, Wright, Nicholson, Kullo and Moyer2020; Gammal, Berenbrok, Empey, & Massart, Reference Gammal, Berenbrok, Empey and Massart2021), (6) high testing costs and inequitable reimbursement (Empey, Pratt, Hoffman, Caudle, & Klein, Reference Empey, Pratt, Hoffman, Caudle and Klein2021), and (7) gaps in end user knowledge and education (Bousman et al., Reference Bousman, Oomen, Jessel, Tampi, Forester, Eyre and Müller2022b; Hayashi & Bousman, Reference Hayashi and Bousman2022; Jameson et al., Reference Jameson, Fylan, Bristow, Sagoo, Dalton, Cardno and McLean2021; Jessel, Al Maruf, Oomen, Arnold, & Bousman, Reference Jessel, Al Maruf, Oomen, Arnold and Bousman2022; Soda et al., Reference Soda, Merner, Small, Torgerson, Muñoz, Austin and Lázaro-Muñoz2023). These hurdles, however, are not insurmountable.

To overcome these hurdles, we recommend: (1) real-world and controlled studies of clinical effectiveness and cost effectiveness to understand clinical and economic benefits of PGx, (2) investment in discovery of additional PGx markers, particularly those related to pharmacodynamic processes, to expand the number of drugs with actionable recommendations, more precisely identify suitable medication options, and improve generalizability in diverse populations, (3) involvement of government or clinical societies in the development and implementation of PGx testing standards via a national or international certification process, (4) partnerships between PGx expert groups (e.g. CPIC) and professional societies that develop clinical practice guidelines to identify opportunities to integrate PGx evidence, (5) prioritization by electronic health/medical record manufactures to improve their product design to seamlessly integrate and present PGx data across systems, (6) encourage health care payors to adopt reimbursement strategies that ensure equitable access to PGx testing, and (7) investment in the development and maintenance of PGx educational materials for healthcare trainees, advanced clinicians, pharmacists, genetic counselors, and the community.

Exemplars of successful and sustainable PGx testing implementation exist (Duarte et al., Reference Duarte, Dalton, Elchynski, Smith, Cicali and Lee2021; Luczak et al., Reference Luczak, Brown, Armbruster, Hundertmark, Brown and Stenehjem2021). In 2004, Cincinnati Children's Hospital Medical Center became the first to offer PGx testing in psychiatry and today, all patients admitted to the inpatient psychiatry service receive PGx testing as part of routine care (Ramsey et al., Reference Ramsey, Prows, Zhang, Saldaña, Sorter, Pestian and Glauser2018). They were joined by other PGx innovators such as St Jude Children's Research Hospital (Hoffman et al., Reference Hoffman, Haidar, Wilkinson, Crews, Baker, Kornegay and Relling2014), Mayo Clinic (Bielinski et al., Reference Bielinski, Olson, Pathak, Weinshilboum, Wang, Lyke and Kullo2014), Vanderbilt University Medical Center (Van Driest et al., Reference Van Driest, Shi, Bowton, Schildcrout, Peterson, Pulley and Roden2014), University of Florida Health Personalized Medicine Program (Cavallari et al., Reference Cavallari, Weitzel, Elsey, Liu, Mosley, Smith and Johnson2017), and several other institutions in the Implementing Genomics in Practice Network (Sperber et al., Reference Sperber, Carpenter, Cavallari, Damschroder, Cooper-DeHoff, Denny and Orlando2017) and Ubiquitous Pharmacogenomics Consortium (Blagec et al., Reference Blagec, Koopmann, Crommentuijn-van Rhenen, Holsappel, van der Wouden, Konta and Samwald2018). These innovators, in part, have created a bridge by which PGx-guided prescribing in psychiatry can successfully cross the chasm. As of March 2023, at least 70 institutions across the globe are implementing PGx-guided prescribing and most include psychiatric indications (CPIC, 2019).

In parallel with the innovative efforts of academic and healthcare institutions, commercial PGx laboratories have also played a role in the clinical implementation of PGx in psychiatry. In 2006, Assurex Health (now Myriad Neuroscience) was founded with patented technology licensed from Cincinnati Children's Hospital Medical Center and Mayo Clinic, marking the launch of the first commercial PGx testing laboratory focused on psychotropic drugs (Assurex Health, 2016). Since then, numerous commercial PGx laboratories have emerged (Bousman & Hopwood, Reference Bousman and Hopwood2016; Maruf et al., Reference Maruf, Fan, Arnold, Müller, Aitchison and Bousman2020) and some medical centers have partnered with these laboratories to provide PGx testing for their patients (RPRD Diagnostics, 2017; St. Catherine Specialty Hospital, n.d.). To date, PGx tests offered by commercial laboratories have been the focus of most PGx clinical trials in psychiatry. For example, 11 of the 13 clinical trials that have examined the efficacy of PGx testing to inform prescribing of antidepressants for patients with depression were supported by commercial PGx testing laboratories (Brown et al., Reference Brown, Stanton, Bharthi, Maruf, Müller and Bousman2022). Pooled results from these clinical trials showed patients receiving PGx-guided antidepressant treatment were 41% (95% CI 15–74%) more likely to achieve symptom remission relative to their counterparts that received treatment as usual (Brown et al., Reference Brown, Stanton, Bharthi, Maruf, Müller and Bousman2022). However, these encouraging trial findings have been tempered by concerns within the academic and clinical community about industry bias, the use of inappropriate comparison groups, and unsatisfactory intervention blinding (Smith & Nemeroff, Reference Smith and Nemeroff2020). Additional concerns have been raised by the FDA related to claims made by commercial PGx testing laboratories, resulting in the issuance of a warning letter to Inova Genomics Laboratory in 2019 (FDA, 2019) and two safety communications cautioning the use of PGx testing results that are not based on FDA-approved product labeling (Ellingrod, Reference Ellingrod2019).

To address the critiques from the academic and clinical community, several investigator-initiated clinical trials have been launched (trial registration numbers: NCT03749629, NCT04623099, NCT04445792, ACTRN12621000181808, ACTRN12621001374853). Moreover, the FDA has become more proactive in the PGx space. In February of 2020, the FDA released and has regularly updated a ‘Table of Pharmacogenetic Associations’ that lists gene–drug pairs that are ‘supported by sufficient scientific evidence’ (Food & Drug Administration, 2021) and more recently released draft guidance for pharmacogenomic data submissions to industry that will replace previous guidance published in 2005 and facilitate the use of PGx data in drug development (FDA, 2023). Unfortunately, only 39 gene–drug pairs of the 106 included in the FDA Table of Associations are included in CPIC guidelines, which has created confusion among PGx implementors and test developers (Pritchard, Patel, Stephens, & McLeod, Reference Pritchard, Patel, Stephens and McLeod2022). However, it is anticipated that concordance between CPIC and the FDA will improve as communication between them is strengthened, and the evidence base continues to grow.

Future growth of PGx in psychiatry

In addition to addressing the implementation hurdles discussed above, the growth of PGx-guided prescribing will, in part, hinge on investment in and integration of knowledge emerging from other pharmaco-omic approaches (e.g. polygenic scoring, pharmacoepigenomics, pharmacometabolomics, pharmacotranscriptomics, pharmacoproteomics, pharmacomicrobiomics). Below, we provide a summary of the knowledge and progress made in these other pharmaco-omic approaches. In the future, we anticipate these approaches in combination with personal and environmental factors will provide an opportunity to facilitate actionable dosing recommendations for additional psychotropic medications and refine current recommendations (Fig. 2).

Figure 2. Conceptual model of precision psychotropic prescribing decision support in the future. PK, pharmacokinetic; PD, pharmacodynamic. Adapted from Maruf and Bousman (Reference Maruf, Bousman, Eyre, Levretsky and Reynolds2021).

The era of big data (e.g. biobanks and genome-wide association studies) has facilitated the development of tools for extracting PGx information from large genomic datasets (McInnes et al., Reference McInnes, Lavertu, Sangkuhl, Klein, Whirl-Carrillo and Altman2021) as well as analytical strategies for interrogating the entire genome rather than individual candidate genes. On such strategy is polygenic scoring in which genetic variation across the genome is mathematically combined to derive a ‘score’ that can be used to predict or stratify a person's probability of experiencing a clinical outcome (e.g. medication response). According to a recent systematic review, 59% (30/51) of all studies using polygenic scores in PGx have been conducted within the area of mental and behavioral diseases (Johnson et al., Reference Johnson, Wilke, Lyle, Kowalec, Jorgensen, Wright and Drögemöller2022). Some of the most notable of these studies have applied this approach to antidepressant (Li, Tian, Hinds, & Team, Reference Li, Tian, Hinds and Team2020; Pain et al., Reference Pain, Hodgson, Trubetskoy, Ripke, Marshe, Adams and GSRD Consortium2022), clozapine (Okhuijsen-Pfeifer et al., Reference Okhuijsen-Pfeifer, van der Horst, Bousman, Lin, van Eijk and Ripke2022; Pardiñas et al., Reference Pardiñas, Kappel, Roberts, Tipple, Shitomi-Jones, King and Walters2023), and lithium (Amare et al., Reference Amare, Thalamuthu, Schubert, Fullerton, Ahmed, Hartmann and Baune2023) response but results are mixed and the variance explained by polygenic scores have been modest (e.g. <10%). Although GWAS and polygenic scoring studies have shown some promise in PGx, clinical implementation is not yet warranted. Going forward, novel approaches for improving the clinical interpretability of polygenic scores are needed. One such approach is pharmagenic enrichment scores, which were developed to provide an indication of an individual's exposure to genomic variants in biological pathways with known drug targets, an approach that may be particularly useful in selection of medications for individuals with treatment-resistant conditions (Reay, Atkins, Carr, Green, & Cairns, Reference Reay, Atkins, Carr, Green and Cairns2020).

Pharmacoepigenomics is a field of study that explores the correlation between epigenetic changes and drug response. Epigenetics refers to heritable and potentially reversible chemical modifications (e.g. methylation, acetylation, phosphorylation) to the genome that can alter gene expression without changing the DNA sequence itself (Dupont, Armant, & Brenner, Reference Dupont, Armant and Brenner2009). It can promote biological impacts that result in a particular phenotype that may be associated with desirable or undesirable response to medications (Micale et al., Reference Micale, Di Bartolomeo, Di Martino, Stark, Dell'Osso, Drago and D'Addario2023). A recent review of antidepressant pharmacoepigenomics found most of the clinical research has focused on methylation of genes encoding the brain derived neurotrophic factor (BDNF), monoamine transporters (SLC6A4, SLC6A2), and serotonin receptors (HTR1A, HTR1B), with promising but mixed results (Hack et al., Reference Hack, Fries, Eyre, Bousman, Singh, Quevedo and Dunlop2019). Most published antidepressant research to date are candidate-gene driven with the exception of one epigenome-wide association study (EWAS) where differential methylation at CpG sites upstream of the CHN2 (encodes beta-chimerin, a protein that is important for cell proliferation) and JAK2 (encodes anus kinase 2, a protein that promotes the growth of cells) transcriptional start site regions were identified as potential predictors of antidepressant response (Ju et al., Reference Ju, Fiori, Belzeaux, Theroux, Chen, Aouabed and Turecki2019). Clinical studies are also encouraging but mixed for antipsychotic pharmacoepigenomics, which have primarily assessed methylation changes in candidate genes linked to the serotonergic (e.g. SLC6A4, HTR1A, HTR2A) and dopaminergic (e.g. DRD2) systems (Ovenden, McGregor, Emsley, & Warnich, Reference Ovenden, McGregor, Emsley and Warnich2018). In a global methylation study, hypomethylation of LINE-1 elements was associated with poor response to risperidone treatment in patients with schizophrenia (Marques et al., Reference Marques, Ota, Santoro, Talarico, Costa, Spindola and Belangero2020). In another whole-genome DNA methylation study, six genes (APIS3, C16orf59, KCNK15, LOC146336, MGC16384, and XRN2) were found to be hypermethylated and were identified as good markers of treatment-induced effects in male schizophrenia patients before and after achieving complete remission (Rukova et al., Reference Rukova, Staneva, Hadjidekova, Stamenov, Milanova and Toncheva2014). Despite these promising findings, pharmacoepigenomic-guided psychotropic prescribing is not ready for clinical implementation, although one can imagine it will contribute to future prescribing decision-making tools. The shift to whole epigenome approaches that not only capture methylation but also other epigenomic mechanisms (e.g. histone modifications) will undoubtedly expedite the identification of informative pharmacoepigenomic profiles and their use in guiding psychotropic treatment.

Pharmacomicrobiomics, the study of how the gut microbiome interacts with medication, is another emerging approach for understanding psychotropic efficacy and tolerability (Cussotto, Clarke, Dinan, & Cryan, Reference Cussotto, Clarke, Dinan and Cryan2019; Generoso, Giridharan, Lee, Macedo, & Barichello, Reference Generoso, Giridharan, Lee, Macedo and Barichello2021; Brown, Bobo, Gall, Muller, & Bousman, Reference Brown, Bobo, Gall, Müller and Bousman2023). Gut microbiota can directly alter absorption, distribution, metabolism, and elimination of psychotropics as well as indirectly modulate host cytochrome (CYP) enzymes (Collins & Patterson, Reference Collins and Patterson2020; Dempsey & Cui, Reference Dempsey and Cui2019). Gut bacteria may also sequester specific medications leading to a change in the microbiome function and in some cases psychotropics can have detrimental or antimicrobial effects on the microbiome (Klünemann et al., Reference Klünemann, Andrejev, Blasche, Mateus, Phapale, Devendran and Patil2021). Moreover, modification of medications by specific bacteria may result in an increase in active metabolites, which could increase response or perpetuate adverse drug reactions (Wan & Zuo, Reference Wan and Zuo2022; Zeng et al., Reference Zeng, Yang, Cao, Gu, Li, Zhang and Cai2021). Therefore, the presence of specific drug-interacting microbial species may serve as a biomarker of efficacy or tolerability to specific medications or guide modification of the microbiota to enrich beneficial species to optimize response to medications (Nita et al., Reference Nita, Ilie, Ciobica, Hritcu, Dobrin, Doroftei and Dobrin2022). Notably, the current evidence in psychiatry is primarily based on preclinical studies (Cussotto et al., Reference Cussotto, Clarke, Dinan and Cryan2019). However, recent work in patients with psychiatric disorders suggests that certain microbiota expression patterns are associated with treatment resistance (Thompson et al., Reference Thompson, Fu, Gandhi, Fowler, Frueh, Weinstein and Madan2023), while other patterns are associated with SSRI (Gao et al., Reference Gao, Tu, Liu, Zhang, Zhang, Jing and Zhang2023) and antipsychotic (Yuan et al., Reference Yuan, Wang, Li, Jiang, Kang, Pang and Song2021) response. Furthermore, clinical trials are ongoing to understand the effect of psychotropic medications on the microbiome (NCT03414151) and the role of the microbiome in psychotropic medication response (NCT05022524). These trials, along with additional clinical studies, are expected to expand our understanding of microbiome–psychotropic drug interactions and highlight opportunities for translation into clinical care.

Pharmacotranscriptomics investigates the associations between variations in the transcriptome (i.e. the complete set of RNA molecules) and the interindividual variation in drug response. Candidate–gene expression studies have found significant associations with SLC6A4 (serotonin transporter), IL1β (interleukin 1β), TNF (tumor necrosis factor alpha), and FKBP5 (FK 506 binding protein 5) genes (reviewed in Belzeaux et al. [Reference Belzeaux, Lin, Ju, Chay, Fiori, Lutz and Turecki2018]). Whole-genome transcriptomic studies have also reported some promising results. For example, interferon regulatory factor 7 (IRF7) was found to be upregulated among citalopram responders in MDD patients (Mamdani et al., Reference Mamdani, Berlim, Beaulieu, Labbe, Merette and Turecki2011). In another study, machine learning approaches identified and independently replicated a set of 13 transcripts that were able to predict MDD patients in symptom remission with an accuracy of 79.4% (Guilloux et al., Reference Guilloux, Bassi, Ding, Walsh, Turecki, Tseng and Sibille2015). A global gene co-expression network analysis found genes related to immune response, acute inflammatory response, and C-X-C motif chemokine ligand 8 (IL8) receptor activity to be associated with antidepressant response using three independent cohorts of patients with depression (Belzeaux et al., Reference Belzeaux, Lin, Ding, Bergon, Ibrahim, Turecki and Sibille2016). Although no transcriptomic markers currently have robust clinical trial evidence, the introduction of new and cheaper RNA sequencing technology will aid in transcriptomic research in psychiatry (reviewed in Maruf & Bousman [Reference Maruf, Bousman, Eyre, Levretsky and Reynolds2021]).

Pharmacoproteomics is an emerging field that investigates changes in the levels of specific proteins in response to medication therapy. In psychiatry, proteomic profiling has focused on specific proteins relevant to cell communication and signaling, inflammation, and cellular growth, and maintenance in response to psychotropic drugs (reviewed in Cassoli, Guest, Santana, & Martins-de-Souza [Reference Cassoli, Guest, Santana and Martins-de-Souza2016]). For example, seven proteins (interleukin-16, fatty acid binding protein, ferritin, C-reactive protein, myoglobin, prolactin, and complement factor H) were found to predict the improvement in positive symptoms, and two proteins (matrix metalloproteinase 2 and insulin) were found to predict improved negative symptoms in patients with schizophrenia treated with antipsychotics (Schwarz, Guest, Steiner, Bogerts, & Bahn, Reference Schwarz, Guest, Steiner, Bogerts and Bahn2012). In the GENDEP (genome-based therapeutic drugs for depression) study cohort, higher C-reactive protein levels were associated with better response with nortriptyline whereas lower levels were associated with better response in patients taking escitalopram (Uher et al., Reference Uher, Maier, Hauser, Marusic, Schmael, Mors and McGuffin2009). To date, no validated proteomic markers are available to aid clinicians in choosing a psychotropic drug for their patient.

Pharmacometabolomics investigates the associations between differences in metabolites and interindividual variances in drug response. Metabolomic profiling can also be used to understand disease pathophysiology and medication mechanism of action. Although promising, pharmacometabolomic studies in psychiatry are very limited. However, the formation of expert groups, such as the Pharmacometabolomics Research Network, suggests we should anticipate findings connecting metabolic profiles to drug response in the future. In fact, the network has already published several studies identifying genetic and metabolic variants relevant to psychotropic drug response (reviewed in Kaddurah-Daouk, Weinshilboum, & Network [Reference Kaddurah-Daouk, Weinshilboum and Network2015]). For example, metabolomic profiling in 290 patients with MDD before and after citalopram/escitalopram treatment showed pathway activation related to tryptophan, tyrosine, and purine metabolism. Interestingly, the study also found differences in metabolites related to gut-microbiota for patients who responded and those who did not after citalopram/escitalopram treatment (Bhattacharyya et al., Reference Bhattacharyya, Ahmed, Arnold, Liu, Luo, Zhu and Kaddurah-Daouk2019), suggesting opportunities for integration of pharmacometabolomics and pharmacomicrobiomics.

Accounting for and integrating non-omic information into medication selection and dosing decision-making tools will also be critical to the growth of PGx-guided prescribing in psychiatry. Desirable and undesirable responses to psychotropic medications also vary by non-genomic information, such as age, sex, renal/hepatic functioning, inflammation, concomitant medications, and lifestyle factors (smoking, diet) (reviewed by Klomp, Manson, Guchelaar, & Swen, [Reference Klomp, Manson, Guchelaar and Swen2020]; Shah & Smith [Reference Shah and Smith2015]). These factors have been shown to modulate expected associations between a genomic-derived factor (e.g. metabolizer status) and medication treatment outcomes through a phenomenon by which a person's genomic-predicted phenotype does not match their clinically observed phenotype. This so-called phenoconversion has been shown to occur frequently in people receiving medications for psychiatric disorders (Mostafa, Kirkpatrick, Byron, & Sheffield, Reference Mostafa, Kirkpatrick, Byron and Sheffield2019; Preskorn et al., Reference Preskorn, Kane, Lobello, Nichols, Fayyad, Buckley and Guico-Pabia2013), but is often not accounted for when interpreting genomic data (Bousman et al., Reference Bousman, Mukerjee, Men, Dorfman, Müller and Thomas2022a). Notably, free phenoconversion tools have been developed to assist laboratories and healthcare providers (Bousman, Wu, Aitchison, & Cheng, Reference Bousman, Wu, Aitchison and Cheng2021; Cicali et al., Reference Cicali, Elchynski, Cook, Houder, Thomas, Smith and Wiisanen2021), although these tools are focused on adjusting genomic-based phenotypes for concomitant medications only. Studies to date have primarily relied on single biomarker interactions with outcomes and do not take multiple factors into account, potentially leading to sub-clinical effects of single markers. The introduction of machine learning algorithms allows for the combination of multiple markers and clinical information to inform medication decisions and predict outcomes by utilizing a greater breadth of information beyond a single marker–drug interaction. A recent review of machine learning applications to antidepressant response showed high predictive accuracies could be obtained but independent external validation was often not performed and when it was, the predictive accuracies of the machine learning algorithms were significantly attenuated (Bobo, Van Ommeren, & Athreya, Reference Bobo, Van Ommeren and Athreya2022). As the evidence strengthens, these tools will evolve to include a wider range of personal and environmental factors that in combination with a greater depth of -omic data will improve psychotropic medication selection and dosing decisions.