BACKGROUND: Massively parallel reporter assays (MPRAs) are an experimental technology for measuring the activity of thousands of candidate regulatory sequences or their variants in parallel, where the activity of individual sequences is measured from pools of sequence-tagged reporter genes. Activity is derived from the ratio of transcribed RNA to input DNA counts of associated tag sequences in each reporter construct, so-called barcodes. Recently, tools specifically designed to analyze MPRA data were developed that attempt to model the count data, accounting for its inherent variation. Of these tools, MPRAnalyze and mpralm are most widely used. MPRAnalyze models barcode counts to estimate the transcription rate of each sequence. While it has increased statistical power and robustness against outliers compared to mpralm, it is slow and has a high false discovery rate. Mpralm, a tool built on the R package Limma, estimates log fold-changes between different sequences. As opposed to MPRAnalyze, it is fast and has a low false discovery rate but is susceptible to outliers and has less statistical power. RESULTS: We propose BCalm, an MPRA analysis framework aimed at addressing the limitations of the existing tools. BCalm is an adaptation of mpralm, but models individual barcode counts instead of aggregating counts per sequence. Leaving out the aggregation step increases statistical power and improves robustness to outliers, while being fast and precise. We show the improved performance over existing methods on both simulated MPRA data and a lentiviral MPRA library of 166,508 target sequences, including 82,258 allelic variants. Further, BCalm adds functionality beyond the existing mpralm package, such as preparing count input files from MPRAsnakeflow, as well as an option to test for sequences with enhancing or repressing activity. Its built-in plotting functionalities allow for easy interpretation of the results. CONCLUSIONS: With BCalm, we provide a new tool for analyzing MPRA data which is robust and accurate on real MPRA datasets. The package is available at https://github.com/kircherlab/BCalm .

MOTIVATION: Missense variants are a frequent class of variation within the coding genome, and some of them cause Mendelian diseases. Despite advances in computational prediction, classifying missense variants into pathogenic or benign remains a major challenge in the context of personalized medicine. Recently, the structure of the human proteome was derived with unprecedented accuracy using the artificial intelligence system AlphaFold2. This raises the question of whether AlphaFold2 wild-type structures can improve the accuracy of computational pathogenicity prediction for missense variants. RESULTS: To address this, we first engineered a set of features for each amino acid from these structures. We then trained a random forest to distinguish between relatively common (proxy-benign) and singleton (proxy-pathogenic) missense variants from gnomAD v3.1. This yielded a novel AlphaFold2-based pathogenicity prediction score, termed AlphScore. Important feature classes used by AlphScore are solvent accessibility, amino acid network related features, features describing the physicochemical environment, and AlphaFold2's quality parameter (predicted local distance difference test). AlphScore alone showed lower performance than existing in silico scores used for missense prediction, such as CADD or REVEL. However, when AlphScore was added to those scores, the performance increased, as measured by the approximation of deep mutational scan data, as well as the prediction of expert-curated missense variants from the ClinVar database. Overall, our data indicate that the integration of AlphaFold2-predicted structures can improve pathogenicity prediction of missense variants. AVAILABILITY AND IMPLEMENTATION: AlphScore, combinations of AlphScore with existing scores, as well as variants used for training and testing are publicly available.

BACKGROUND: Cis-regulatory regions (CRRs) are non-coding regions of the DNA that fine control the spatio-temporal pattern of transcription; they are involved in a wide range of pivotal processes such as the development of specific cell-lines/tissues and the dynamic cell response to physiological stimuli. Recent studies showed that genetic variants occurring in CRRs are strongly correlated with pathogenicity or deleteriousness. Considering the central role of CRRs in the regulation of physiological and pathological conditions, the correct identification of CRRs and of their tissue-specific activity status through Machine Learning methods plays a major role in dissecting the impact of genetic variants on human diseases. Unfortunately, the problem is still open, though some promising results have been already reported by (deep) machine-learning based methods that predict active promoters and enhancers in specific tissues or cell lines by encoding epigenetic or spectral features directly extracted from DNA sequences. RESULTS: We present the experiments we performed to compare two Deep Neural Networks, a Feed-Forward Neural Network model working on epigenomic features, and a Convolutional Neural Network model working only on genomic sequence, targeted to the identification of enhancer- and promoter-activity in specific cell lines. While performing experiments to understand how the experimental setup influences the prediction performance of the methods, we particularly focused on (1) automatic model selection performed by Bayesian optimization and (2) exploring different data rebalancing setups for reducing negative unbalancing effects. CONCLUSIONS: Results show that (1) automatic model selection by Bayesian optimization improves the quality of the learner; (2) data rebalancing considerably impacts the prediction performance of the models; test set rebalancing may provide over-optimistic results, and should therefore be cautiously applied; (3) despite working on sequence data, convolutional models obtain performance close to those of feed forward models working on epigenomic information, which suggests that also sequence data carries informative content for CRR-activity prediction. We therefore suggest combining both models/data types in future works.

Targeted sequencing of genomic regions is a cost- and time-efficient approach for screening patient cohorts. We present a fast and efficient workflow to analyze highly imbalanced, targeted next-generation sequencing data generated using molecular inversion probe (MIP) capture. Our Snakemake pipeline performs sample demultiplexing, overlap paired-end merging, alignment, MIP-arm trimming, variant calling, coverage analysis and report generation. Further, we support the analysis of probes specifically designed to capture certain structural variants and can assign sex using Y-chromosome-unique probes. In a user-friendly HTML report, we summarize all these results including covered, incomplete or missing regions, called variants and their predicted effects. We developed and tested our pipeline using the hemophilia A & B MIP design from the "My Life, Our Future" initiative. HemoMIPs is available as an open-source tool on GitHub at: https://github.com/kircherlab/hemoMIPs.

MOTIVATION: The conversion of the raw intensities obtained from next-generation sequencing platforms into nucleotide sequences with well-calibrated quality scores is a critical step in the generation of good sequence data. While recent model-based approaches can yield highly accurate calls, they require a substantial amount of processing time and/or computational resources. We previously introduced Ibis, a fast and accurate basecaller for the Illumina platform. We have continued active development of Ibis to take into account developments in the Illumina technology, as well as to make Ibis fully open source. RESULTS: We introduce here freeIbis, which offers significant improvements in sequence accuracy owing to the use of a novel multiclass support vector machine (SVM) algorithm. Sequence quality scores are now calibrated based on empirically observed scores, thus providing a high correlation to their respective error rates. These improvements result in downstream advantages including improved genotyping accuracy. AVAILABILITY AND IMPLEMENTATION: FreeIbis is freely available for use under the GPL (http://bioinf.eva.mpg.de/freeibis/). It requires a Python interpreter and a C++ compiler. Tailored versions of LIBOCAS and LIBLINEAR are distributed along with the package.