Dataset Information

Prediction of lithium response using genomic data.

ABSTRACT: Predicting lithium response prior to treatment could both expedite therapy and avoid exposure to side effects. Since lithium responsiveness may be heritable, its predictability based on genomic data is of interest. We thus evaluate the degree to which lithium response can be predicted with a machine learning (ML) approach using genomic data. Using the largest existing genomic dataset in the lithium response literature (n = 2210 across 14 international sites; 29% responders), we evaluated the degree to which lithium response could be predicted based on 47,465 genotyped single nucleotide polymorphisms using a supervised ML approach. Under appropriate cross-validation procedures, lithium response could be predicted to above-chance levels in two constituent sites (Halifax, Cohen's kappa 0.15, 95% confidence interval, CI [0.07, 0.24]; and Würzburg, kappa 0.2 [0.1, 0.3]). Variants with shared importance in these models showed over-representation of postsynaptic membrane related genes. Lithium response was not predictable in the pooled dataset (kappa 0.02 [- 0.01, 0.04]), although non-trivial performance was achieved within a restricted dataset including only those patients followed prospectively (kappa 0.09 [0.04, 0.14]). Genomic classification of lithium response remains a promising but difficult task. Classification performance could potentially be improved by further harmonization of data collection procedures.

SUBMITTER: Stone W

PROVIDER: S-EPMC7806976 | biostudies-literature | 2021 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Prediction of lithium response using genomic data.

Stone William W Nunes Abraham A Akiyama Kazufumi K Akula Nirmala N Ardau Raffaella R Aubry Jean-Michel JM Backlund Lena L Bauer Michael M Bellivier Frank F Cervantes Pablo P Chen Hsi-Chung HC Chillotti Caterina C Cruceanu Cristiana C Dayer Alexandre A Degenhardt Franziska F Del Zompo Maria M Forstner Andreas J AJ Frye Mark M Fullerton Janice M JM Grigoroiu-Serbanescu Maria M Grof Paul P Hashimoto Ryota R Hou Liping L Jiménez Esther E Kato Tadafumi T Kelsoe John J Kittel-Schneider Sarah S Kuo Po-Hsiu PH Kusumi Ichiro I Lavebratt Catharina C Manchia Mirko M Martinsson Lina L Mattheisen Manuel M McMahon Francis J FJ Millischer Vincent V Mitchell Philip B PB Nöthen Markus M MM O'Donovan Claire C Ozaki Norio N Pisanu Claudia C Reif Andreas A Rietschel Marcella M Rouleau Guy G Rybakowski Janusz J Schalling Martin M Schofield Peter R PR Schulze Thomas G TG Severino Giovanni G Squassina Alessio A Veeh Julia J Vieta Eduard E Trappenberg Thomas T Alda Martin M

Scientific reports 20210113 1

Predicting lithium response prior to treatment could both expedite therapy and avoid exposure to side effects. Since lithium responsiveness may be heritable, its predictability based on genomic data is of interest. We thus evaluate the degree to which lithium response can be predicted with a machine learning (ML) approach using genomic data. Using the largest existing genomic dataset in the lithium response literature (n = 2210 across 14 international sites; 29% responders), we evaluated the deg ...[more]

PMID: 33441847

Similar Datasets

Project description:BackgroundLithium, one of the few effective treatments for bipolar depression (BPD), has been hypothesized to work by enhancing serotonergic transmission. Despite preclinical evidence, it is unknown whether lithium acts via the serotonergic system. Here we examined the potential of serotonin transporter (5-HTT) or serotonin 1A receptor (5-HT1A) pre-treatment binding to predict lithium treatment response and remission. We hypothesized that lower pre-treatment 5-HTT and higher pre-treatment 5-HT1A binding would predict better clinical response. Additional analyses investigated group differences between BPD and healthy controls and the relationship between change in binding pre- to post-treatment and clinical response. Twenty-seven medication-free patients with BPD currently in a depressive episode received positron emission tomography (PET) scans using 5-HTT tracer [11C]DASB, a subset also received a PET scan using 5-HT1A tracer [11C]-CUMI-101 before and after 8 weeks of lithium monotherapy. Metabolite-corrected arterial input functions were used to estimate binding potential, proportional to receptor availability. Fourteen patients with BPD with both [11C]DASB and [11C]-CUMI-101 pre-treatment scans and 8 weeks of post-treatment clinical scores were included in the prediction analysis examining the potential of either pre-treatment 5-HTT or 5-HT1A or the combination of both to predict post-treatment clinical scores.ResultsWe found lower pre-treatment 5-HTT binding (p = 0.003) and lower 5-HT1A binding (p = 0.035) were both significantly associated with improved clinical response. Pre-treatment 5-HTT predicted remission with 71% accuracy (77% specificity, 60% sensitivity), while 5-HT1A binding was able to predict remission with 85% accuracy (87% sensitivity, 80% specificity). The combined prediction analysis using both 5-HTT and 5-HT1A was able to predict remission with 84.6% accuracy (87.5% specificity, 60% sensitivity). Additional analyses BPD and controls pre- or post-treatment, and the change in binding were not significant and unrelated to treatment response (p > 0.05).ConclusionsOur findings suggest that while lithium may not act directly via 5-HTT or 5-HT1A to ameliorate depressive symptoms, pre-treatment binding may be a potential biomarker for successful treatment of BPD with lithium.Clinical trial registrationPET and MRI Brain Imaging of Bipolar Disorder Identifier: NCT01880957; URL: https://clinicaltrials.gov/ct2/show/NCT01880957.

Dataset Information

Prediction of lithium response using genomic data.

Publications

Prediction of lithium response using genomic data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets