Pan's Algorithm Research Articles

Precision Neoantigen Discovery Using Large-Scale Immunopeptidomes and Composite Modeling of MHC Peptide Presentation

Major histocompatibility complex (MHC)–bound peptides that originate from tumor-specific genetic alterations, known as neoantigens, are an important class of anticancer therapeutic targets. Accurately predicting peptide presentation by MHC complexes is a key aspect of discovering therapeutically relevant neoantigens. Technological improvements in mass spectrometry–based immunopeptidomics and advanced modeling techniques have vastly improved MHC presentation prediction over the past 2 decades. However, improvement in the accuracy of prediction algorithms is needed for clinical applications like the development of personalized cancer vaccines, the discovery of biomarkers for response to immunotherapies, and the quantification of autoimmune risk in gene therapies. Toward this end, we generated allele-specific immunopeptidomics data using 25 monoallelic cell lines and created Systematic Human Leukocyte Antigen (HLA) Epitope Ranking Pan Algorithm (SHERPA), a pan-allelic MHC-peptide algorithm for predicting MHC-peptide binding and presentation. In contrast to previously published large-scale monoallelic data, we used an HLA-null K562 parental cell line and a stable transfection of HLA allele to better emulate native presentation. Our dataset includes five previously unprofiled alleles that expand MHC diversity in the training data and extend allelic coverage in underprofiled populations. To improve generalizability, SHERPA systematically integrates 128 monoallelic and 384 multiallelic samples with publicly available immunoproteomics data and binding assay data. Using this dataset, we developed two features that empirically estimate the propensities of genes and specific regions within gene bodies to engender immunopeptides to represent antigen processing. Using a composite model constructed with gradient boosting decision trees, multiallelic deconvolution, and 2.15 million peptides encompassing 167 alleles, we achieved a 1.44-fold improvement of positive predictive value compared with existing tools when evaluated on independent monoallelic datasets and a 1.17-fold improvement when evaluating on tumor samples. With a high degree of accuracy, SHERPA has the potential to enable precision neoantigen discovery for future clinical applications.

Clinical Fusion for Real-Time Complex QRS Pattern Detection in Wearable ECG Using the Pan-Tompkins Algorithm

This scientific paper presents a novel approach of real-time signal analysis in electrocardiogram (ECG) monitoring systems, focusing on the integration of device design,algorithm implementation for accurate measurement and interpretation of heart activity. The proposed system leverages a low-cost framework, employing a microcontroller and Arduino programming language for raw ECG data acquisition, while utilizing the AD8232 sensor and ESP8266 Node MCU for continuous patient monitoring. The acquired data is processed, stored, and analyzed using the Pan-Tompkins algorithm, which effectively filters and analyzes heart signals, including noise reduction and QRS complex detection. Two case studies involving a healthy individual and a patient with Myocarditis were conducted to demonstrate the effectiveness of the system. The integration of device design and algorithm development in ECG analysis is emphasized, highlighting the affordability, wearability, and potential for continuous monitoring and early detection of heart conditions. By successfully mitigating noise-related challenges, the implementation of the Pan algorithm enables accurate signal analysis. This interdisciplinary research contributes to the advancement of ECG interpretation and underscores the significance of clinical fusion between designed systems and applied algorithms on real cases. The performance of two Pan-Tompkins based QRS complex detection algorithms was systematically analyzed, offering valuable insights for their reasonable utilization.

Sensitivity and Positive Prediction of Secured Electrocardiograph (ECG) Transmission using Fully Homomorphic Encryption Technique (FHE)

Health connected is a technology that links medical devices, telecommunications and security techniques. It empowers patients to be observed and treated remotely from their homes. Patient’s healthcare records with a connected healthcare system should be stored securely before transmitted for further investigation and interpretation. Electrocardiogram (ECG) is the clinical method utilized to screen heart execution and utilized for the detection of various arrhythmias. For diagnostic purposes, individuals with a background of heart diseases have long records of ECGs, which results in the requirement of a large amount of storage space and labor. Hence, there is a requirement for a system that involves digital signal processing and signal security so that the spared information is made sure about at one spot and an only authentic individual can see and utilize this ECG signal for additional findings. This study presents a set of security solutions that can be deployed in a connected healthcare territory, which includes the fully homomorphic encryption (FHE) techniques used to secure the ECG signals. The study helps the medical provider to record ECG signals confidentially and to prevent mistreatment. The study focuses on Pan and Tompkins algorithm methods for the detection of the ECG Signal. As a result, the output of the Pan and Tompkins algorithm for ECG signal processing with the FHE technique shows a sensitivity of 92.59% and a positive prediction of 90.00%.

Abstract 1898: Accurate modeling of antigen processing and MHC peptide presentation using large-scale immunopeptidomes and a novel machine learning framework

Abstract Neoantigens, which are antigens specific to cancer cells, can be harnessed to develop precision immunotherapies, such as personalized cancer vaccines, and prognostic biomarkers for checkpoint blockade inhibition. Next generation sequencing technologies have enabled comprehensive profiling of putative neoantigens by interrogating the tumor exome and transcriptome, but accurate prediction of peptides presented by MHC complexes remains a significant challenge. We present here Systematic HLA Epitope Ranking Pan Algorithm (SHERPA™) that addresses this critical need. SHERPA comprises highly sensitive, accurate and pan-allelic MHC-peptide (MHCp) binding and presentation prediction models, that were built using a multi-pronged strategy. First, we generated a large-scale, high-quality HLA ligandome using approximately 75 stably transfected mono-allelic K562 cell lines. Intact MHCp complexes were immunoprecipitated using W6/32 antibody and profiled using LC/MS-MS. Second, we trained models that predict both MHCp binding and presentation. Briefly, MHCp binding was modeled using the amino acid sequences of the ligand and the binding pocket of the cognate allele. MHCp presentation, which encompasses in vivo antigen processing, was modeled using multiple features including the expression level of the source protein, proteasomal cleavage, and two novel features representing presentation propensities of genes and regions within gene bodies. Third, we expanded the scale and scope of our in-house dataset using a large curated repository of publicly available mono- and multi-allelic datasets resulting in > 160 alleles and > 1.6 million peptides. Integrating data from diverse cell line and tissue types improved the generalizability of our models, a critically important aspect when applying our models to patient samples. Finally, we implemented a model-based deconvolution of multi-allelic datasets to generate pseudo mono-allelic data, and developed an integrative machine learning architecture to model our expanded HLA-ligandome. We evaluated the performance of our binding and prediction models on 10% held-out mono-allelic test data from multiple cell line sources. The precision at various recall values of both binding and prediction models was markedly higher than NetMHCPan 4.0, and the positive predictive values were 0.59 and 0.73 respectively, significantly higher compared to NetMHCpan 4.0 (PPV = 0.38). Additionally, a strong concordance of raw and predicted motifs for alleles excluded from training data indicated a robust pan-allelic performance. When evaluated on 12 tissue samples profiled in-house, the SHERPA presentation model had a consistently high recall (90%) compared to NetMHCpan 4.0 (63%). This trend holds true on external immunopeptidomics datasets from tumor samples. In summary, SHERPA enables precision neoantigen discovery. Citation Format: Rachel Marty Pyke, Dattatreya Mellacheruvu, Steven Dea, Charles Abbott, Nick Phillips, Sejal Desai, Rena McClory, Steven Ketelaars, Pia Kvistborg, John West, Richard Chen, Sean Michael Boyle. Accurate modeling of antigen processing and MHC peptide presentation using large-scale immunopeptidomes and a novel machine learning framework [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1898.

Precision Neoantigen Discovery Using Large-scale Immunopeptidomes and Composite Modeling of MHC Peptide Presentation

Major histocompatibility complex (MHC)-bound peptides that originate from tumor-specific genetic alterations, known as neoantigens, are an important class of anticancer therapeutic targets. Accurately predicting peptide presentation by MHC complexes is a key aspect of discovering therapeutically relevant neoantigens. Technological improvements in mass-spectrometry-based immunopeptidomics and advanced modeling techniques have vastly improved MHC presentation prediction over the past two decades. However, improvement in the sensitivity and specificity of prediction algorithms is needed for clinical applications such as the development of personalized cancer vaccines, the discovery of biomarkers for response to checkpoint blockade, and the quantification of autoimmune risk in gene therapies. Toward this end, we generated allele-specific immunopeptidomics data using 25 monoallelic cell lines and created Systematic HLA Epitope Ranking Pan Algorithm (SHERPA), a pan-allelic MHC-peptide algorithm for predicting MHC-peptide binding and presentation. In contrast to previously published large-scale monoallelic data, we used an HLA-null K562 parental cell line and a stable transfection of HLA alleles to better emulate native presentation. Our dataset includes five previously unprofiled alleles that expand MHC-binding pocket diversity in the training data and extend allelic coverage in under profiled populations. To improve generalizability, SHERPA systematically integrates 128 monoallelic and 384 multiallelic samples with publicly available immunoproteomics data and binding assay data. Using this dataset, we developed two features that empirically estimate the propensities of genes and specific regions within gene bodies to engender immunopeptides to represent antigen processing. Using a composite model constructed with gradient boosting decision trees, multiallelic deconvolution, and 2.15 million peptides encompassing 167 alleles, we achieved a 1.44-fold improvement of positive predictive value compared with existing tools when evaluated on independent monoallelic datasets and a 1.15-fold improvement when evaluating on tumor samples. With a high degree of accuracy, SHERPA has the potential to enable precision neoantigen discovery for future clinical applications.

Integrated IoT intelligent system for the automatic detection of cardiac variability.

Detection of abnormal cardiac events during clinical examination is a matter of chances, as such events may not happen at that precise moment. We therefore propose the implementation and evaluation of a mobile based system that allows a real-time detection of cardiovascular problems related to heart-rate variability. Our approach is to integrate an Internet of Things eHealth kit based on Arduino and validated algorithms for heart rate variability to build a low-cost, reliable and scalable solution. 12 healthy users have evaluated the system in different scenarios to assess the best performing algorithm and the best windowing interval. Finally, a mobile system based on an Android application which integrated the Pan and Tompkins algorithm with a 20 seconds windowing and a module to retrieve real-time electrocardiography through a Bluetooth interface was implemented and assessed.

Univariate real root isolation in an extension field and applications

We present algorithmic, complexity and implementation results for the problem of isolating the real roots of a univariate polynomial in Bα∈L[y], where L=Q(α) is a simple algebraic extension of the rational numbers.We revisit two approaches for the problem. In the first approach, using resultant computations, we perform a reduction to a polynomial with integer coefficients and we deduce a bound of O˜B(N8) for isolating the real roots of Bα, where N is an upper bound on all the quantities (degree and bitsize) of the input polynomials. The bound becomes O˜B(N7) if we use Pan's algorithm for isolating the real roots. In the second approach we isolate the real roots working directly on the polynomial of the input. We compute improved separation bounds for the roots and we prove that they are optimal, under mild assumptions. For isolating the real roots we consider a modified Sturm algorithm, and a modified version of descartes' algorithm. For the former we prove a Boolean complexity bound of O˜B(N12) and for the latter a bound of O˜B(N5). We present aggregate separation bounds and complexity results for isolating the real roots of all polynomials Bαk, when αk runs over all the real conjugates of α. We show that we can isolate the real roots of all polynomials in O˜B(N5). Finally, we implemented the algorithms in C as part of the core library of MATHEMATICA and we illustrate their efficiency over various data sets.

Tighter estimates for ϵ-nets for disks

The geometric hitting set problem is one of the basic geometric combinatorial optimization problems: given a set P of points, and a set D of geometric objects in the plane, the goal is to compute a small-sized subset of P that hits all objects in D. In 1994, Bronnimann and Goodrich [5] made an important connection of this problem to the size of fundamental combinatorial structures called ϵ-nets, showing that small-sized ϵ-nets imply approximation algorithms with correspondingly small approximation ratios. Very recently, Agarwal and Pan [2] showed that their scheme can be implemented in near-linear time for disks in the plane. Altogether this gives O(1)-factor approximation algorithms in O˜(n) time for hitting sets for disks in the plane.This constant factor depends on the sizes of ϵ-nets for disks; unfortunately, the current state-of-the-art bounds are large – at least 24/ϵ and most likely larger than 40/ϵ. Thus the approximation factor of the Agarwal and Pan algorithm ends up being more than 40. The best lower-bound is 2/ϵ, which follows from the Pach–Woeginger construction [32] for halfplanes in two dimensions. Thus there is a large gap between the best-known upper and lower bounds. Besides being of independent interest, finding precise bounds is important since this immediately implies an improved linear-time algorithm for the hitting-set problem.The main goal of this paper is to improve the upper-bound to 13.4/ϵ for disks in the plane. The proof is constructive, giving a simple algorithm that uses only Delaunay triangulations. We have implemented the algorithm, which is available as a public open-source module. Experimental results show that the sizes of ϵ-nets for a variety of data-sets are lower, around 9/ϵ.

From approximate factorization to root isolation with application to cylindrical algebraic decomposition

We present an algorithm for isolating all roots of an arbitrary complex polynomial p that also works in the presence of multiple roots provided that (1) the number of distinct roots is given as part of the input and (2) the algorithm can ask for arbitrarily good approximations of the coefficients of p. The algorithm outputs pairwise disjoint disks each containing one of the distinct roots of p and the multiplicity of the root contained in the disk. The algorithm uses approximate factorization as a subroutine. For the case where Pan's algorithm (Pan, 2002) is used for the factorization, we derive complexity bounds for the problems of isolating and refining all roots, which are stated in terms of the geometric locations of the roots only. Specializing the latter bounds to a polynomial of degree d and with integer coefficients of bitsize less than τ, we show that O˜(d3+d2τ+dκ) bit operations are sufficient to compute isolating disks of size less than 2−κ for all roots of p, where κ is an arbitrary positive integer.In addition, we apply our root isolation algorithm to a recent algorithm for computing the topology of a real planar algebraic curve specified as the zero set of a bivariate integer polynomial and for isolating the real solutions of a bivariate polynomial system. For polynomials of degree n and bitsize τ, we improve the currently best running time from O˜(n9τ+n8τ2) (deterministic) to O˜(n6+n5τ) (randomized) for topology computation and from O˜(n8+n7τ) (deterministic) to O˜(n6+n5τ) (randomized) for solving bivariate systems.

New fast prediction mode selection algorithm in H.264/AVC frame

In order to gain the optimal mode,H.264 adopts Rate Distortion Optimization(RDO) technique to compute the cost of every mode for each macroblock,but it increases computational complexity.A precision mode selection algorithm in fast frame was proposed which combined PAN algorithm proposed by Pan.Firstly,the intra prediction's type were decided by whether a macroblock was smooth.Secondly,the PAN algorithm was improved and a new algorithm based on the statistical properties of adjacent blocks was proposed.The experimental results show that the method mentioned in this paper reduces the total encoding time with negligible loss in Peak Signal-to-Noise Ratio(PSNR) and a slightly increased bitrate,compared with full search algorithm and PAN algorithm.

A New Vertical Diffusion Package with an Explicit Treatment of Entrainment Processes

Abstract This paper proposes a revised vertical diffusion package with a nonlocal turbulent mixing coefficient in the planetary boundary layer (PBL). Based on the study of Noh et al. and accumulated results of the behavior of the Hong and Pan algorithm, a revised vertical diffusion algorithm that is suitable for weather forecasting and climate prediction models is developed. The major ingredient of the revision is the inclusion of an explicit treatment of entrainment processes at the top of the PBL. The new diffusion package is called the Yonsei University PBL (YSU PBL). In a one-dimensional offline test framework, the revised scheme is found to improve several features compared with the Hong and Pan implementation. The YSU PBL increases boundary layer mixing in the thermally induced free convection regime and decreases it in the mechanically induced forced convection regime, which alleviates the well-known problems in the Medium-Range Forecast (MRF) PBL. Excessive mixing in the mixed layer in the presence of strong winds is resolved. Overly rapid growth of the PBL in the case of the Hong and Pan is also rectified. The scheme has been successfully implemented in the Weather Research and Forecast model producing a more realistic structure of the PBL and its development. In a case study of a frontal tornado outbreak, it is found that some systematic biases of the large-scale features such as an afternoon cold bias at 850 hPa in the MRF PBL are resolved. Consequently, the new scheme does a better job in reproducing the convective inhibition. Because the convective inhibition is accurately predicted, widespread light precipitation ahead of a front, in the case of the MRF PBL, is reduced. In the frontal region, the YSU PBL scheme improves some characteristics, such as a double line of intense convection. This is because the boundary layer from the YSU PBL scheme remains less diluted by entrainment leaving more fuel for severe convection when the front triggers it.