PandoGen: Generating complete instances of future SARS-CoV-2 sequences using Deep Learning.

One of the challenges in a viral pandemic is the emergence of novel variants with different phenotypical characteristics. An ability to forecast future viral individuals at the sequence level enables advance preparation by characterizing the sequences and closing vulnerabilities in current preventative and therapeutic methods. In this article, we explore, in the context of a viral pandemic, the problem of generating complete instances of undiscovered viral protein sequences, which have a high likelihood of being discovered in the future using protein language models. Current approaches to training these models fit model parameters to a known sequence set, which does not suit pandemic forecasting as future sequences differ from known sequences in some respects. To address this, we develop a novel method, called PandoGen, to train protein language models towards the pandemic protein forecasting task. PandoGen combines techniques such as synthetic data generation, conditional sequence generation, and reward-based learning, enabling the model to forecast future sequences, with a high propensity to spread. Applying our method to modeling the SARS-CoV-2 Spike protein sequence, we find empirically that our model forecasts twice as many novel sequences with five times the case counts compared to a model that is 30× larger. Our method forecasts unseen lineages months in advance, whereas models 4× and 30× larger forecast almost no new lineages. When trained on data available up to a month before the onset of important Variants of Concern, our method consistently forecasts sequences belonging to those variants within tight sequence budgets.

HELLO: improved neural network architectures and methodologies for small variant calling.

BACKGROUND: Modern Next Generation- and Third Generation- Sequencing methods such as Illumina and PacBio Circular Consensus Sequencing platforms provide accurate sequencing data. Parallel developments in Deep Learning have enabled the application of Deep Neural Networks to variant calling, surpassing the accuracy of classical approaches in many settings. DeepVariant, arguably the most popular among such methods, transforms the problem of variant calling into one of image recognition where a Deep Neural Network analyzes sequencing data that is formatted as images, achieving high accuracy. In this paper, we explore an alternative approach to designing Deep Neural Networks for variant calling, where we use meticulously designed Deep Neural Network architectures and customized variant inference functions that account for the underlying nature of sequencing data instead of converting the problem to one of image recognition. RESULTS: Results from 27 whole-genome variant calling experiments spanning Illumina, PacBio and hybrid Illumina-PacBio settings suggest that our method allows vastly smaller Deep Neural Networks to outperform the Inception-v3 architecture used in DeepVariant for indel and substitution-type variant calls. For example, our method reduces the number of indel call errors by up to 18%, 55% and 65% for Illumina, PacBio and hybrid Illumina-PacBio variant calling respectively, compared to a similarly trained DeepVariant pipeline. In these cases, our models are between 7 and 14 times smaller. CONCLUSIONS: We believe that the improved accuracy and problem-specific customization of our models will enable more accurate pipelines and further method development in the field. HELLO is available at https://github.com/anands-repo/hello.

Multimode smartphone biosensing: the transmission, reflection, and intensity spectral (TRI)-analyzer.

We demonstrate a smartphone-integrated handheld detection instrument capable of utilizing the internal rear-facing camera as a high-resolution spectrometer for measuring the colorimetric absorption spectrum, fluorescence emission spectrum, and resonant reflection spectrum from a microfluidic cartridge inserted into the measurement light path. Under user selection, the instrument gathers light from either the white "flash" LED of the smartphone or an integrated green laser diode to direct illumination into a liquid test sample or onto a photonic crystal biosensor. Light emerging from each type of assay is gathered via optical fiber and passed through a diffraction grating placed directly over the smartphone camera to generate spectra from the assay when an image is collected. Each sensing modality is associated with a unique configuration of a microfluidic "stick" containing a linear array of liquid chambers that are swiped through the instrument while the smartphone captures video and the software automatically selects spectra representative of each compartment. The system is demonstrated for representative assays in the field of point-of-care (POC) maternal and infant health: an ELISA assay for the fetal fibronectin protein used as an indicator for pre-term birth and a fluorescent assay for phenylalanine as an indicator for phenylketonuria. In each case, the TRI-analyzer is capable of achieving limits of detection that are comparable to those obtained for the same assay measured with a conventional laboratory microplate reader, demonstrating the flexibility of the system to serve as a platform for rapid, simple translation of existing commercially available biosensing assays to a POC setting.

Characterization of drug authenticity using thin-layer chromatography imaging with a mobile phone.

Thin-layer chromatography (TLC) has a myriad of separation applications in chemistry, biology, and pharmacology due to its simplicity and low cost. While benchtop laboratory sample application and detection systems for TLC provide accurate quantitation of TLC spot positions and densities, there are many applications where inexpensive and portable instruments would greatly expand the applicability of the technology. In this work, we demonstrate identity verification and concentration determination of pharmaceutical compounds via TLC using a custom 3D-printed cradle that interfaces with an ordinary mobile phone. The cradle holds the mobile phone's internal, rear-facing camera in a fixed position relative to a UV lamp and a TLC plate that includes a phosphor in the stationary phase. Analysis of photographs thus reveals the locations and intensities of principal spots of UV--absorbing drugs. Automated image analysis software determines the center location and density of dark spots, which, using integrated calibration spots of known drug compounds and concentrations, can be used to determine if a drug has been diluted or substituted. Two independent image processing approaches have been developed that may be selected based upon the processing capabilities of the smartphone. Each approach is able to discern 5% drug concentration differences. Using single-component solutions of nevirapine, amodiaquine, and paracetamol that have been manually applied, the mobile phone-based detection instrument provides measurements that are equivalent to those obtained with a commercially available lab-based desktop TLC densitometer.