PARE: A framework for removal of confounding effects from any distance-based dimension reduction method.

Dimension reduction tools preserving similarity and graph structure such as t-SNE and UMAP can capture complex biological patterns in high-dimensional data. However, these tools typically are not designed to separate effects of interest from unwanted effects due to confounders. We introduce the partial embedding (PARE) framework, which enables removal of confounders from any distance-based dimension reduction method. We then develop partial t-SNE and partial UMAP and apply these methods to genomic and neuroimaging data. For lower-dimensional visualization, our results show that the PARE framework can remove batch effects in single-cell sequencing data as well as separate clinical and technical variability in neuroimaging measures. We demonstrate that the PARE framework extends dimension reduction methods to highlight biological patterns of interest while effectively removing confounding effects.

Familial Mediterranean Fever and multiple sclerosis treated with ocrelizumab: Case report.

BACKGROUND: Familial Mediterranean Fever (FMF) is associated with increased risk of multiple sclerosis (MS). Optimal treatment of patients with comorbid FMF and MS remains uncertain. CASE: A 28-year-old woman with FMF, treated with colchicine, had symptomatic onset of relapsing remitting MS following four simultaneous vaccines. MRI brain with a 7-Tesla magnet demonstrated several areas of leptomeningeal enhancement with predominant linear, spread/fill and rare nodular patterns. Central vein signs were present in supratentorial white matter lesions. She received four cycles of ocrelizumab and achieved no evidence of disease activity (NEDA-3) at 20 months' follow up. DISCUSSION: FMF with incident CNS demyelinating disease demonstrated neuroimaging features typical for classic RRMS including the central vein sign and leptomeningeal enhancement. Treatment with B-cell depleting therapy for FMF-MS led to clinical stability and symptomatic improvement at 20 months' follow up. We add to the sparse literature characterizing the course of FMF as a genetic risk factor for CNS demyelinating disease, demonstrating pathognomonic imaging features of MS on 7 T imaging and treatment efficacy with B-cell depletion.

Deconfounded Dimension Reduction via Partial Embeddings.

Dimension reduction tools preserving similarity and graph structure such as t-SNE and UMAP can capture complex biological patterns in high-dimensional data. However, these tools typically are not designed to separate effects of interest from unwanted effects due to confounders. We introduce the partial embedding (PARE) framework, which enables removal of confounders from any distance-based dimension reduction method. We then develop partial t-SNE and partial UMAP and apply these methods to genomic and neuroimaging data. Our results show that the PARE framework can remove batch effects in single-cell sequencing data as well as separate clinical and technical variability in neuroimaging measures. We demonstrate that the PARE framework extends dimension reduction methods to highlight biological patterns of interest while effectively removing confounding effects.

Exploring the effect of glatiramer acetate on cerebral gray matter atrophy in multiple sclerosis.

BACKGROUND AND PURPOSE: Cerebral gray matter (GM) atrophy is a proposed measure of neuroprotection in multiple sclerosis (MS). Glatiramer acetate (GA) limits clinical relapses, MRI lesions, and whole brain atrophy in relapsing-remitting MS (RRMS). The effect of GA on GM atrophy remains unclear. We assessed GM atrophy in patients with RRMS starting GA therapy in comparison to a cohort of patients with clinically benign RRMS (BMS). DESIGN/METHODS: We studied 14 patients at GA start [age (mean ± SD) 44.2 ± 7.0 years, disease duration (DD) 7.2 ± 6.4 years, Expanded Disability Status Scale score (EDSS) (median,IQR) 1.0,2.0] and 6 patients with BMS [age 43.0 ± 6.1 years, DD 18.1 ± 8.4 years, EDSS 0.5,1.0]. Brain MRI was obtained at baseline and one year later (both groups) and two years later in all patients in the GA group except one who was lost to follow-up. Semi-automated algorithms assessed cerebral T2 hyperintense lesion volume (T2LV), white matter fraction (WMF), GM fraction (GMF), and brain parenchymal fraction (BPF). The exact Wilcoxon-Mann-Whitney test compared the groups. The Wilcoxon signed rank test assessed longitudinal changes within groups. RESULTS: During the first year, MRI changes did not differ significantly between groups (p > 0.15). Within the BMS group, WMF and BPF decreased during the first year (p = 0.03). Within the GA group, there was no significant change in MRI measures during each annual period (p > 0.05). Over two years, the GA group had a significant increase in T2LV and decrease in WMF (p < 0.05), while GMF and BPF remained stable (p > 0.05). MRI changes in brain volumes (GMF or WMF) in the first year in the GA group were not significantly different from those in the BMS group (p > 0.5). CONCLUSIONS: In this pilot study with a small sample size, patients with RRMS started on GA did not show significant GM or whole brain atrophy over 2 years, resembling MS patients with a clinically benign disease course.

Embedding Scientific Explanations Into Storybooks Impacts Children's Scientific Discourse and Learning.

Children's understanding of unobservable scientific entities largely depends on testimony from others, especially through parental explanations that highlight the mechanism underlying a scientific entity. Mechanistic explanations are particularly helpful in promoting children's conceptual understanding, yet they are relatively rare in parent-child conversations. The current study aimed to increase parent-child use of mechanistic conversation by modeling this language in a storybook about the mechanism of electrical circuits. We also examined whether an increase in mechanistic conversation was associated with science learning outcomes, measured at both the dyadic- and child-level. In the current study, parents and their 4- to 5-year-old children (N = 60) were randomly assigned to read a book containing mechanistic explanations (n = 32) or one containing non-mechanistic explanations (n = 28). After reading the book together, parent-child joint understanding of electricity's mechanism was tested by asking the dyad to assemble electrical components of a circuit toy so that a light would turn on. Finally, child science learning outcomes were examined by asking children to assemble a novel circuit toy and answer comprehension questions to gauge their understanding of electricity's mechanism. Results indicate that dyads who read storybooks containing mechanistic explanations were (1) more successful at completing the circuit (putting the pieces together to make the light turn on) and (2) used more mechanistic language than dyads assigned to the non-mechanistic condition. Children in the mechanistic condition also had better learning outcomes, but only if they engaged in more mechanistic discourse with their parent. We discuss these results using a social interactionist framework to highlight the role of input and interaction for learning. We also highlight how these results implicate everyday routines such as book reading in supporting children's scientific discourse and understanding.

The Bayesian risk estimate at onset (BREMSO) correlates with cognitive and physical disability in patients with early multiple sclerosis.

BACKGROUND: Prevention of long-term disability is the goal of therapeutic intervention in Relapsing Remitting MS (RRMS). The Bayesian Risk Estimate for MS at Onset (BREMSO) gives an individual risk score predicting disease evolution into Secondary Progressive MS (SPMS). We investigated whether BREMSO correlates with physical disability, cognitive dysfunction, and regional brain atrophy early in MS. METHODS: One hundred RRMS patients with at least two years of follow-up were enrolled. BREMSO score as well as Symbol Digit Modalities Test (SDMT) and Multiple Sclerosis Severity Score (MSSS), Timed 25-Foot Walk Test (T25-FW) and 9-Hole Peg Test (9-HPT), were assessed. Intracranial volume (ICV), subcortical gray matter structures and corpus callosum (CC) were automatically segmented on MRI images and their volumes measured. RESULTS: BREMSO score correlated negatively with SDMT at visit1 (ß = -0.33, p = 0.019), visit2 (ß = -0.34, p = 0.017) and visit3 (ß = -0.34, p = 0.014), and positively with MSSS at visit1 (r = 0.38, p = 0.006), visit2 (r = 0.47, p < 0.0001) and visit3 (r = 0.42, p = 0.002), but not with T25-FW and 9-HPT. BREMSO negatively correlated with CC volume at baseline (p < 0.03). No correlations were found with ICV and subcortical gray matter. CONCLUSIONS: BREMSO score at onset correlated with physical disability (MSSS), cognitive function (SDMT) and CC volume measurements in patients with early MS.