A guide to machine learning for biologists.

The expanding scale and inherent complexity of biological data have encouraged a growing use of machine learning in biology to build informative and predictive models of the underlying biological processes. All machine learning techniques fit models to data; however, the specific methods are quite varied and can at first glance seem bewildering. In this Review, we aim to provide readers with a gentle introduction to a few key machine learning techniques, including the most recently developed and widely used techniques involving deep neural networks. We describe how different techniques may be suited to specific types of biological data, and also discuss some best practices and points to consider when one is embarking on experiments involving machine learning. Some emerging directions in machine learning methodology are also discussed.

New Brain Tumor Entities Emerge from Molecular Classification of CNS-PNETs.

Primitive neuroectodermal tumors of the central nervous system (CNS-PNETs) are highly aggressive, poorly differentiated embryonal tumors occurring predominantly in young children but also affecting adolescents and adults. Herein, we demonstrate that a significant proportion of institutionally diagnosed CNS-PNETs display molecular profiles indistinguishable from those of various other well-defined CNS tumor entities, facilitating diagnosis and appropriate therapy for patients with these tumors. From the remaining fraction of CNS-PNETs, we identify four new CNS tumor entities, each associated with a recurrent genetic alteration and distinct histopathological and clinical features. These new molecular entities, designated "CNS neuroblastoma with FOXR2 activation (CNS NB-FOXR2)," "CNS Ewing sarcoma family tumor with CIC alteration (CNS EFT-CIC)," "CNS high-grade neuroepithelial tumor with MN1 alteration (CNS HGNET-MN1)," and "CNS high-grade neuroepithelial tumor with BCOR alteration (CNS HGNET-BCOR)," will enable meaningful clinical trials and the development of therapeutic strategies for patients affected by poorly differentiated CNS tumors.

Connect four with glioblastoma stem cell factors.

Hierarchical cell state models, wherein a few stem-like tumor-propagating cells repopulate the tumor after therapy, are often invoked in cancer. Suvà et al. demonstrate a plastic developmental hierarchy in glioma cell populations by characterizing the epigenetic states of phenotypically distinct cells and identifying four factors sufficient to reprogram differentiated cells into a tumorigenic stem-like state.

Connectome-based modelling of neurodegenerative diseases: towards precision medicine and mechanistic insight.

Neurodegenerative diseases are the most common cause of dementia. Although their underlying molecular pathologies have been identified, there is substantial heterogeneity in the patterns of progressive brain alterations across and within these diseases. Recent advances in neuroimaging methods have revealed that pathological proteins accumulate along specific macroscale brain networks, implicating the network architecture of the brain in the system-level pathophysiology of neurodegenerative diseases. However, the extent to which 'network-based neurodegeneration' applies across the wide range of neurodegenerative disorders remains unclear. Here, we discuss the state-of-the-art of neuroimaging-based connectomics for the mapping and prediction of neurodegenerative processes. We review findings supporting brain networks as passive conduits through which pathological proteins spread. As an alternative view, we also discuss complementary work suggesting that network alterations actively modulate the spreading of pathological proteins between connected brain regions. We conclude this Perspective by proposing an integrative framework in which connectome-based models can be advanced along three dimensions of innovation: incorporating parameters that modulate propagation behaviour on the basis of measurable biological features; building patient-tailored models that use individual-level information and allowing model parameters to interact dynamically over time. We discuss promises and pitfalls of these strategies for improving disease insights and moving towards precision medicine.

Hypermutation of the inactive X chromosome is a frequent event in cancer.

Mutation is a fundamental process in tumorigenesis. However, the degree to which the rate of somatic mutation varies across the human genome and the mechanistic basis underlying this variation remain to be fully elucidated. Here, we performed a cross-cancer comparison of 402 whole genomes comprising a diverse set of childhood and adult tumors, including both solid and hematopoietic malignancies. Surprisingly, we found that the inactive X chromosome of many female cancer genomes accumulates on average twice and up to four times as many somatic mutations per megabase, as compared to the individual autosomes. Whole-genome sequencing of clonally expanded hematopoietic stem/progenitor cells (HSPCs) from healthy individuals and a premalignant myelodysplastic syndrome (MDS) sample revealed no X chromosome hypermutation. Our data suggest that hypermutation of the inactive X chromosome is an early and frequent feature of tumorigenesis resulting from DNA replication stress in aberrantly proliferating cells.

The Power of Human Cancer Genetics as Revealed by Low-Grade Gliomas.

The human brain contains a vast number of cells and shows extraordinary cellular diversity to facilitate the many cognitive and automatic commands governing our bodily functions. This complexity arises partly from large-scale structural variations in the genome, evolutionary processes to increase brain size, function, and cognition. Not surprisingly given recent technical advances, low-grade gliomas (LGGs), which arise from the glia (the most abundant cell type in the brain), have undergone a recent revolution in their classification and therapy, especially in the pediatric setting. Next-generation sequencing has uncovered previously unappreciated diverse LGG entities, unraveling genetic subgroups and multiple molecular alterations and altered pathways, including many amenable to therapeutic targeting. In this article we review these novel entities, in which oncogenic processes show striking age-related neuroanatomical specificity (highlighting their close interplay with development); the opportunities they provide for targeted therapies, some of which are already practiced at the bedside; and the challenges of implementing molecular pathology in the clinic.

Genome sequencing of pediatric medulloblastoma links catastrophic DNA rearrangements with TP53 mutations.

Genomic rearrangements are thought to occur progressively during tumor development. Recent findings, however, suggest an alternative mechanism, involving massive chromosome rearrangements in a one-step catastrophic event termed chromothripsis. We report the whole-genome sequencing-based analysis of a Sonic-Hedgehog medulloblastoma (SHH-MB) brain tumor from a patient with a germline TP53 mutation (Li-Fraumeni syndrome), uncovering massive, complex chromosome rearrangements. Integrating TP53 status with microarray and deep sequencing-based DNA rearrangement data in additional patients reveals a striking association between TP53 mutation and chromothripsis in SHH-MBs. Analysis of additional tumor entities substantiates a link between TP53 mutation and chromothripsis, and indicates a context-specific role for p53 in catastrophic DNA rearrangements. Among these, we observed a strong association between somatic TP53 mutations and chromothripsis in acute myeloid leukemia. These findings connect p53 status and chromothripsis in specific tumor types, providing a genetic basis for understanding particularly aggressive subtypes of cancer.

Improved protein structure prediction using potentials from deep learning.

Protein structure prediction can be used to determine the three-dimensional shape of a protein from its amino acid sequence1. This problem is of fundamental importance as the structure of a protein largely determines its function2; however, protein structures can be difficult to determine experimentally. Considerable progress has recently been made by leveraging genetic information. It is possible to infer which amino acid residues are in contact by analysing covariation in homologous sequences, which aids in the prediction of protein structures3. Here we show that we can train a neural network to make accurate predictions of the distances between pairs of residues, which convey more information about the structure than contact predictions. Using this information, we construct a potential of mean force4 that can accurately describe the shape of a protein. We find that the resulting potential can be optimized by a simple gradient descent algorithm to generate structures without complex sampling procedures. The resulting system, named AlphaFold, achieves high accuracy, even for sequences with fewer homologous sequences. In the recent Critical Assessment of Protein Structure Prediction5 (CASP13)-a blind assessment of the state of the field-AlphaFold created high-accuracy structures (with template modelling (TM) scores6 of 0.7 or higher) for 24 out of 43 free modelling domains, whereas the next best method, which used sampling and contact information, achieved such accuracy for only 14 out of 43 domains. AlphaFold represents a considerable advance in protein-structure prediction. We expect this increased accuracy to enable insights into the function and malfunction of proteins, especially in cases for which no structures for homologous proteins have been experimentally determined7.

Germline Elongator mutations in Sonic Hedgehog medulloblastoma.

Cancer genomics has revealed many genes and core molecular processes that contribute to human malignancies, but the genetic and molecular bases of many rare cancers remains unclear. Genetic predisposition accounts for 5 to 10% of cancer diagnoses in children1,2, and genetic events that cooperate with known somatic driver events are poorly understood. Pathogenic germline variants in established cancer predisposition genes have been recently identified in 5% of patients with the malignant brain tumour medulloblastoma3. Here, by analysing all protein-coding genes, we identify and replicate rare germline loss-of-function variants across ELP1 in 14% of paediatric patients with the medulloblastoma subgroup Sonic Hedgehog (MBSHH). ELP1 was the most common medulloblastoma predisposition gene and increased the prevalence of genetic predisposition to 40% among paediatric patients with MBSHH. Parent-offspring and pedigree analyses identified two families with a history of paediatric medulloblastoma. ELP1-associated medulloblastomas were restricted to the molecular SHHα subtype4 and characterized by universal biallelic inactivation of ELP1 owing to somatic loss of chromosome arm 9q. Most ELP1-associated medulloblastomas also exhibited somatic alterations in PTCH1, which suggests that germline ELP1 loss-of-function variants predispose individuals to tumour development in combination with constitutive activation of SHH signalling. ELP1 is the largest subunit of the evolutionarily conserved Elongator complex, which catalyses translational elongation through tRNA modifications at the wobble (U34) position5,6. Tumours from patients with ELP1-associated MBSHH were characterized by a destabilized Elongator complex, loss of Elongator-dependent tRNA modifications, codon-dependent translational reprogramming, and induction of the unfolded protein response, consistent with loss of protein homeostasis due to Elongator deficiency in model systems7-9. Thus, genetic predisposition to proteome instability may be a determinant in the pathogenesis of paediatric brain cancers. These results support investigation of the role of protein homeostasis in other cancer types and potential for therapeutic interference.

Deep learning for the PSIPRED Protein Analysis Workbench.

The PSIRED Workbench is a long established and popular bioinformatics web service offering a wide range of machine learning based analyses for characterizing protein structure and function. In this paper we provide an update of the recent additions and developments to the webserver, with a focus on new Deep Learning based methods. We briefly discuss some trends in server usage since the publication of AlphaFold2 and we give an overview of some upcoming developments for the service. The PSIPRED Workbench is available at http://bioinf.cs.ucl.ac.uk/psipred.

Conumee 2.0: enhanced copy-number variation analysis from DNA methylation arrays for humans and mice.

MOTIVATION: Copy-number variations (CNVs) are common genetic alterations in cancer and their detection may impact tumor classification and therapeutic decisions. However, detection of clinically relevant large and focal CNVs remains challenging when sample material or resources are limited. This has motivated us to create a software tool to infer CNVs from DNA methylation arrays which are often generated as part of clinical routines and in research settings. RESULTS: We present our R package, conumee 2.0, that combines tangent normalization, an adjustable genomic binning heuristic, and weighted circular binary segmentation to utilize DNA methylation arrays for CNV analysis and mitigate technical biases and batch effects. Segmentation results were validated in a lung squamous cell carcinoma dataset from TCGA (n = 367 samples) by comparison to segmentations derived from genotyping arrays (Pearson's correlation coefficient of 0.91). We further introduce a segmented block bootstrapping approach to detect focal alternations that achieved 60.9% sensitivity and 98.6% specificity for deletions affecting CDKN2A/B (60.0% and 96.9% for RB1, respectively) in a low-grade glioma cohort from TCGA (n = 239 samples). Finally, our tool provides functionality to detect and summarize CNVs across large sample cohorts. AVAILABILITY AND IMPLEMENTATION: Conumee 2.0 is available under open-source license at: https://github.com/hovestadtlab/conumee2.

Uncovering the distinct macro-scale anatomy of dysexecutive and behavioural degenerative diseases.

There is a longstanding ambiguity regarding the clinical diagnosis of dementia syndromes predominantly targeting executive functions versus behaviour and personality. This is due to an incomplete understanding of the macro-scale anatomy underlying these symptomatologies, a partial overlap in clinical features and the fact that both phenotypes can emerge from the same pathology and vice versa. We collected data from a patient cohort of which 52 had dysexecutive Alzheimer's disease, 30 had behavioural variant frontotemporal dementia (bvFTD), seven met clinical criteria for bvFTD but had Alzheimer's disease pathology (behavioural Alzheimer's disease) and 28 had amnestic Alzheimer's disease. We first assessed group-wise differences in clinical and cognitive features and patterns of fluorodeoxyglucose (FDG) PET hypometabolism. We then performed a spectral decomposition of covariance between FDG-PET images to yield latent patterns of relative hypometabolism unbiased by diagnostic classification, which are referred to as 'eigenbrains'. These eigenbrains were subsequently linked to clinical and cognitive data and meta-analytic topics from a large external database of neuroimaging studies reflecting a wide range of mental functions. Finally, we performed a data-driven exploratory linear discriminant analysis to perform eigenbrain-based multiclass diagnostic predictions. Dysexecutive Alzheimer's disease and bvFTD patients were the youngest at symptom onset, followed by behavioural Alzheimer's disease, then amnestic Alzheimer's disease. Dysexecutive Alzheimer's disease patients had worse cognitive performance on nearly all cognitive domains compared with other groups, except verbal fluency which was equally impaired in dysexecutive Alzheimer's disease and bvFTD. Hypometabolism was observed in heteromodal cortices in dysexecutive Alzheimer's disease, temporo-parietal areas in amnestic Alzheimer's disease and frontotemporal areas in bvFTD and behavioural Alzheimer's disease. The unbiased spectral decomposition analysis revealed that relative hypometabolism in heteromodal cortices was associated with worse dysexecutive symptomatology and a lower likelihood of presenting with behaviour/personality problems, whereas relative hypometabolism in frontotemporal areas was associated with a higher likelihood of presenting with behaviour/personality problems but did not correlate with most cognitive measures. The linear discriminant analysis yielded an accuracy of 82.1% in predicting diagnostic category and did not misclassify any dysexecutive Alzheimer's disease patient for behavioural Alzheimer's disease and vice versa. Our results strongly suggest a double dissociation in that distinct macro-scale underpinnings underlie predominant dysexecutive versus personality/behavioural symptomatology in dementia syndromes. This has important implications for the implementation of criteria to diagnose and distinguish these diseases and supports the use of data-driven techniques to inform the classification of neurodegenerative diseases.

Synthesizing images of tau pathology from cross-modal neuroimaging using deep learning.

Given the prevalence of dementia and the development of pathology-specific disease-modifying therapies, high-value biomarker strategies to inform medical decision-making are critical. In vivo tau-PET is an ideal target as a biomarker for Alzheimer's disease diagnosis and treatment outcome measure. However, tau-PET is not currently widely accessible to patients compared to other neuroimaging methods. In this study, we present a convolutional neural network (CNN) model that imputes tau-PET images from more widely available cross-modality imaging inputs. Participants (n = 1192) with brain T1-weighted MRI (T1w), fluorodeoxyglucose (FDG)-PET, amyloid-PET and tau-PET were included. We found that a CNN model can impute tau-PET images with high accuracy, the highest being for the FDG-based model followed by amyloid-PET and T1w. In testing implications of artificial intelligence-imputed tau-PET, only the FDG-based model showed a significant improvement of performance in classifying tau positivity and diagnostic groups compared to the original input data, suggesting that application of the model could enhance the utility of the metabolic images. The interpretability experiment revealed that the FDG- and T1w-based models utilized the non-local input from physically remote regions of interest to estimate the tau-PET, but this was not the case for the Pittsburgh compound B-based model. This implies that the model can learn the distinct biological relationship between FDG-PET, T1w and tau-PET from the relationship between amyloid-PET and tau-PET. Our study suggests that extending neuroimaging's use with artificial intelligence to predict protein specific pathologies has great potential to inform emerging care models.

Ultrafast end-to-end protein structure prediction enables high-throughput exploration of uncharacterized proteins.

Deep learning-based prediction of protein structure usually begins by constructing a multiple sequence alignment (MSA) containing homologs of the target protein. The most successful approaches combine large feature sets derived from MSAs, and considerable computational effort is spent deriving these input features. We present a method that greatly reduces the amount of preprocessing required for a target MSA, while producing main chain coordinates as a direct output of a deep neural network. The network makes use of just three recurrent networks and a stack of residual convolutional layers, making the predictor very fast to run, and easy to install and use. Our approach constructs a directly learned representation of the sequences in an MSA, starting from a one-hot encoding of the sequences. When supplemented with an approximate precision matrix, the learned representation can be used to produce structural models of comparable or greater accuracy as compared to our original DMPfold method, while requiring less than a second to produce a typical model. This level of accuracy and speed allows very large-scale three-dimensional modeling of proteins on minimal hardware, and we demonstrate this by producing models for over 1.3 million uncharacterized regions of proteins extracted from the BFD sequence clusters. After constructing an initial set of approximate models, we select a confident subset of over 30,000 models for further refinement and analysis, revealing putative novel protein folds. We also provide updated models for over 5,000 Pfam families studied in the original DMPfold paper.

Capmatinib is an effective treatment for MET-fusion driven pediatric high-grade glioma and synergizes with radiotherapy.

BACKGROUND: Pediatric-type diffuse high-grade glioma (pHGG) is the most frequent malignant brain tumor in children and can be subclassified into multiple entities. Fusion genes activating the MET receptor tyrosine kinase often occur in infant-type hemispheric glioma (IHG) but also in other pHGG and are associated with devastating morbidity and mortality. METHODS: To identify new treatment options, we established and characterized two novel orthotopic mouse models harboring distinct MET fusions. These included an immunocompetent, murine allograft model and patient-derived orthotopic xenografts (PDOX) from a MET-fusion IHG patient who failed conventional therapy and targeted therapy with cabozantinib. With these models, we analyzed the efficacy and pharmacokinetic properties of three MET inhibitors, capmatinib, crizotinib and cabozantinib, alone or combined with radiotherapy. RESULTS: Capmatinib showed superior brain pharmacokinetic properties and greater in vitro and in vivo efficacy than cabozantinib or crizotinib in both models. The PDOX models recapitulated the poor efficacy of cabozantinib experienced by the patient. In contrast, capmatinib extended survival and induced long-term progression-free survival when combined with radiotherapy in two complementary mouse models. Capmatinib treatment increased radiation-induced DNA double-strand breaks and delayed their repair. CONCLUSIONS: We comprehensively investigated the combination of MET inhibition and radiotherapy as a novel treatment option for MET-driven pHGG. Our seminal preclinical data package includes pharmacokinetic characterization, recapitulation of clinical outcomes, coinciding results from multiple complementing in vivo studies, and insights into molecular mechanism underlying increased efficacy. Taken together, we demonstrate the groundbreaking efficacy of capmatinib and radiation as a highly promising concept for future clinical trials.

Clinically unfavorable transcriptome subtypes of non-WNT/non-SHH medulloblastomas are associated with a predominance in proliferating and progenitor-like cell subpopulations.

The non-WNT/non-SHH (Grp3/Grp4) medulloblastomas (MBs) include eight second-generation subgroups (SGS; I-VIII) each with distinct molecular and clinical characteristics. Recently, we also identified two prognostically relevant transcriptome subtypes within each SGS MB, which are associated with unique gene expression signatures and signaling pathways. These prognostic subsets may be in connection to the intra-tumoral cell landscape that underlies SGS MB clinical-molecular diversity. Here, we performed a deconvolution analysis of the Grp3/Grp4 MB bulk RNA profiles using the previously identified single-cell RNA-seq reference dataset and focusing on variability in the cellular composition of SGS MB. RNA deconvolution analysis of the Grp3/Grp4 MB disclosed the subgroup-specific neoplastic cell subpopulations. Neuronally differentiated axodendritic GP3-C1 and glutamatergic GP4-C1 subpopulations were distributed within Grp3- and Grp4-associated SGS MB, respectively. Progenitor GP3-B2 subpopulation was prominent in aggressive SGS II MB, whereas photoreceptor/visual perception GP3/4-C2 cell content was typical for SGS III/IV MB. The current study also revealed significant variability in the proportions of cell subpopulations between clinically relevant SGS MB transcriptome subtypes, where unfavorable cohorts were enriched with cell cycle and progenitor-like cell subpopulations and, vice versa, favorable subtypes were composed of neuronally differentiated cell fractions predominantly. A higher than median proportion of proliferating and progenitor cell subpopulations conferred the shortest survival of the Grp3 and Grp 4 MB, and similar survival associations were identified for all SGS MB except SGS IV MB. In summary, the recently identified clinically relevant Grp3/Grp4 MB transcriptome subtypes are composed of different cell populations. Future studies should aim to validate the prognostic and therapeutic role of the identified Grp3/Grp4 MB inter-tumoral cellular heterogeneity. The application of the single-cell techniques on each SGS MB separately could help to clarify the clinical significance of subgroup-specific variability in tumor cell content and its relation with prognostic transcriptome signatures identified before.

Clinicoradiological and neuropathological evaluation of primary progressive aphasia.

BACKGROUND: Primary progressive aphasia (PPA) defines a group of neurodegenerative disorders characterised by language decline. Three PPA variants correlate with distinct underlying pathologies: semantic variant PPA (svPPA) with transactive response DNA-binding protein of 43 kD (TDP-43) proteinopathy, agrammatic variant PPA (agPPA) with tau deposition and logopenic variant PPA (lvPPA) with Alzheimer's disease (AD). Our objectives were to differentiate PPA variants using clinical and neuroimaging features, assess progression and evaluate structural MRI and a novel 18-F fluorodeoxyglucose positron emission tomography (FDG-PET) image decomposition machine learning algorithm for neuropathology prediction. METHODS: We analysed 82 autopsied patients diagnosed with PPA from 1998 to 2022. Clinical histories, language characteristics, neuropsychological results and brain imaging were reviewed. A machine learning framework using a k-nearest neighbours classifier assessed FDG-PET scans from 45 patients compared with a large reference database. RESULTS: PPA variant distribution: 35 lvPPA (80% AD), 28 agPPA (89% tauopathy) and 18 svPPA (72% frontotemporal lobar degeneration-TAR DNA-binding protein (FTLD-TDP)). Apraxia of speech was associated with 4R-tauopathy in agPPA, while pure agrammatic PPA without apraxia was linked to 3R-tauopathy. Longitudinal data revealed language dysfunction remained the predominant deficit for patients with lvPPA, agPPA evolved to corticobasal or progressive supranuclear palsy syndrome (64%) and svPPA progressed to behavioural variant frontotemporal dementia (44%). agPPA-4R-tauopathy exhibited limited pre-supplementary motor area atrophy, lvPPA-AD displayed temporal atrophy extending to the superior temporal sulcus and svPPA-FTLD-TDP had severe temporal pole atrophy. The FDG-PET-based machine learning algorithm accurately predicted clinical diagnoses and underlying pathologies. CONCLUSIONS: Distinguishing 3R-taupathy and 4R-tauopathy in agPPA may rely on apraxia of speech presence. Additional linguistic and clinical features can aid neuropathology prediction. Our data-driven brain metabolism decomposition approach effectively predicts underlying neuropathology.

LOGGIC/FIREFLY-2: a phase 3, randomized trial of tovorafenib vs. chemotherapy in pediatric and young adult patients with newly diagnosed low-grade glioma harboring an activating RAF alteration.

BACKGROUND: Pediatric low-grade glioma (pLGG) is essentially a single pathway disease, with most tumors driven by genomic alterations affecting the mitogen-activated protein kinase/ERK (MAPK) pathway, predominantly KIAA1549::BRAF fusions and BRAF V600E mutations. This makes pLGG an ideal candidate for MAPK pathway-targeted treatments. The type I BRAF inhibitor, dabrafenib, in combination with the MEK inhibitor, trametinib, has been approved by the United States Food and Drug Administration for the systemic treatment of BRAF V600E-mutated pLGG. However, this combination is not approved for the treatment of patients with tumors harboring BRAF fusions as type I RAF inhibitors are ineffective in this setting and may paradoxically enhance tumor growth. The type II RAF inhibitor, tovorafenib (formerly DAY101, TAK-580, MLN2480), has shown promising activity and good tolerability in patients with BRAF-altered pLGG in the phase 2 FIREFLY-1 study, with an objective response rate (ORR) per Response Assessment in Neuro-Oncology high-grade glioma (RANO-HGG) criteria of 67%. Tumor response was independent of histologic subtype, BRAF alteration type (fusion vs. mutation), number of prior lines of therapy, and prior MAPK-pathway inhibitor use. METHODS: LOGGIC/FIREFLY-2 is a two-arm, randomized, open-label, multicenter, global, phase 3 trial to evaluate the efficacy, safety, and tolerability of tovorafenib monotherapy vs. current standard of care (SoC) chemotherapy in patients < 25 years of age with pLGG harboring an activating RAF alteration who require first-line systemic therapy. Patients are randomized 1:1 to either tovorafenib, administered once weekly at 420 mg/m2 (not to exceed 600 mg), or investigator's choice of prespecified SoC chemotherapy regimens. The primary objective is to compare ORR between the two treatment arms, as assessed by independent review per RANO-LGG criteria. Secondary objectives include comparisons of progression-free survival, duration of response, safety, neurologic function, and clinical benefit rate. DISCUSSION: The promising tovorafenib activity data, CNS-penetration properties, strong scientific rationale combined with the manageable tolerability and safety profile seen in patients with pLGG led to the SIOPe-BTG-LGG working group to nominate tovorafenib for comparison with SoC chemotherapy in this first-line phase 3 trial. The efficacy, safety, and functional response data generated from the trial may define a new SoC treatment for newly diagnosed pLGG. TRIAL REGISTRATION: ClinicalTrials.gov: NCT05566795. Registered on October 4, 2022.

Pineal anlage tumor: clinical and diagnostic features, and rationales for treatment.

PURPOSE: To provide a treatment-focused review and develop basic treatment guidelines for patients diagnosed with pineal anlage tumor (PAT). METHODS: Prospectively collected data of three patients with pineal anlage tumor from Germany was combined with clinical details and treatment information from 17 published cases. RESULTS: Overall, 20 cases of PAT were identified (3 not previously reported German cases, 17 cases from published reports). Age at diagnosis ranged from 0.3 to 35.0 (median: 3.2 ± 7.8) years. All but three cases were diagnosed before the age of three years. For three cases, metastatic disease at initial staging was described. All patients underwent tumor surgery (gross-total resection: 9, subtotal resection/biopsy: 9, extent of resection unknown: 2). 15/20 patients were alive at last follow-up. Median follow-up for 10/15 surviving patients with available follow-up and treatment data was 2.4 years (0.3-6.5). Relapse was reported for 3 patients within 0.8 years after diagnosis. Five patients died, 3 after relapse and 2 from early postoperative complications. Two-year-progression-free- and -overall survival were 65.2 ± 12.7% and 49.2 ± 18.2%, respectively. All 4 patients who received intensive chemotherapy including high-dose chemotherapy combined with radiotherapy (2 focal, 2 craniospinal [CSI]) had no recurrence. Focal radiotherapy- and CSI-free survival rates in 13 evaluable patients were 46.2% (6/13) and 61.5% (8/13), respectively. CONCLUSION: PAT is an aggressive disease mostly affecting young children. Therefore, adjuvant therapy using intensive chemotherapy and considering radiotherapy appears to comprise an appropriate treatment strategy. Reporting further cases is crucial to evaluate distinct treatment strategies.