Conversion Artifacts Research Articles

Framework to enable and test conversational assistant for APIs and RPAs

AbstractIn the realm of business automation, conversational assistants are emerging as the primary method for making automation software accessible to users in various business sectors. Access to automation primarily occurs through application programming interface (APIs) and robotic process automation (RPAs). To effectively convert APIs and RPAs into chatbots on a larger scale, it is crucial to establish an automated process for generating data and training models that can recognize user intentions, identify questions for conversational slot filling, and provide recommendations for subsequent actions. In this paper, we present a technique for enhancing and generating natural language conversational artifacts from API specifications using large language models (LLMs). The goal is to utilize LLMs in the “build” phase to assist humans in creating skills for digital assistants. As a result, the system does not need to rely on LLMs during conversations with business users, leading to efficient deployment. Along with enabling digital assistants, our system employs LLMs as proxies to simulate human interaction and automatically evaluate the digital assistant's performance. Experimental results highlight the effectiveness of our proposed approach. Our system is deployed in the IBM Watson Orchestrate product for general availability.

Intan Technologies integrated circuits can produce analog-to-digital conversion artifacts that affect neural signal acquisition

Objective. Intan Technologies’ integrated circuits (ICs) are valuable tools for neurophysiological data acquisition, providing signal amplification, filtering, and digitization from many channels (up to 64 channels/chip) at high sampling rates (up to 30 kSPS) within a compact package (⩽9× 7 mm). However, we found that the analog-to-digital converters (ADCs) in the Intan RHD2000 series ICs can produce artifacts in recorded signals. Here, we examine the effects of these ADC artifacts on neural signal quality and describe a method to detect them in recorded data. Approach. We identified two types of ADC artifacts produced by Intan ICs: 1) jumps, resulting from missing output codes, and 2) flatlines, resulting from overrepresented output codes. We identified ADC artifacts in neural recordings acquired with Intan RHD2000 ICs and tested the repeated performance of 17 ICs in vitro. With the on-chip digital-signal-processing disabled, we detected the ADC artifacts in each test recording by examining the distribution of unfiltered ADC output codes. Main Results. We found larger ADC artifacts in recordings using the Intan RHX data acquisition software versions 3.0–3.2, which did not run the necessary ADC calibration command when the inputs to the Intan recording controller were rescanned. This has been corrected in the Intan RHX software version 3.3. We found that the ADC calibration routine significantly reduced, but did not fully eliminate, the occurrence and size of ADC artifacts as compared with recordings acquired when the calibration routine was not run (p < 0.0001). When the ADC calibration routine was run, we found that the artifacts produced by each ADC were consistent over time, enabling us to sort ICs by performance. Significance. Our findings call attention to the importance of evaluating signal quality when acquiring electrophysiological data using Intan Technologies ICs and offer a method for detecting ADC artifacts in recorded data.

Building Conversational Artifacts to Enable Digital Assistant for APIs and RPAs

In the realm of business automation, digital assistants/chatbots are emerging as the primary method for making automation software accessible to users in various business sectors. Access to automation primarily occurs through APIs and RPAs. To effectively convert APIs and RPAs into chatbots on a larger scale, it is crucial to establish an automated process for generating data and training models that can recognize user intentions, identify questions for conversational slot filling, and provide recommendations for subsequent actions. In this paper, we present a technique for enhancing and generating natural language conversational artifacts from API specifications using large language models (LLMs). The goal is to utilize LLMs in the "build" phase to assist humans in creating skills for digital assistants. As a result, the system doesn't need to rely on LLMs during conversations with business users, leading to efficient deployment. Experimental results highlight the effectiveness of our proposed approach. Our system is deployed in the IBM Watson Orchestrate product for general availability.

Abstract 960: Improved cfDNA methylation profiling through correction of misrepaired jagged-ends

Abstract Background: Profiling cfDNA methylation enables early cancer detection and classification, but challenges arise from "jagged ends" (JEs) for ~90% of cfDNA molecules, as well as for FFPE samples and partially degraded genomic DNA. Repair of these jagged ends introduces erroneous methylation signals, especially in shorter cfDNA fragments enriched for tumor derived molecules. Separately, DNA conversions during 5mC profiling hamper accurate simultaneous ctDNA genotyping of somatic mutations. To meet these needs, we developed a method preserving double-stranded cfDNA molecules without error-prone end-repair at jagged ends. This enables superior methylation profiling including hemimethylation detection, and simultaneous ctDNA SNV genotyping. Methods: We present JEEPERS (Jagged-End Error Polishing of Enzymatically misRepaired Sequences), a novel method for correcting errors in double-stranded library preparation for methylation profiling. JEEPERS detects and quantifies enzymatic repair errors at JEs, correcting them in silico. Leveraging duplex UMIs, it utilizes complementary strands and sibling reads from cfDNA families without jagged ends. The method exploits the 5' to 3' polarity of ER-associated errors and JE length profiles for correction. JEEPERS introduces anchor CpGs for precise correction of rare signals and enables accurate SNV calling at non CpG sites, addressing the challenge of differentiating true SNVs from conversion artifacts leading to C&amp;gt;T/G&amp;gt;A changes in sequencing data. Results: Analyzing 248 cfDNA samples (139 NSCLC patients, 109 healthy controls) by methylation profiling (EM-Seq) and ctDNA mutation genotyping (CAPP-Seq), JEEPERS correction efficiently addressed distortions, even at low tumor concentrations. Limiting dilution experiments demonstrated JEEPERS achieving a LOD95 of ~0.2% with 95% specificity (n=300). Despite ~50% lower molecular depth, JEEPERS-corrected cfDNA EM-Seq enabled accurate simultaneous methylation profiling and ctDNA genotyping, correlating strongly with CAPP-Seq genotypes (R=0.998, n=30). As expected, mutant ctDNA molecules were shorter than wild-type in both CAPP-Seq and EM-Seq data. cfDNA methylation patterns varied by oncogene vs. tumor suppressor status; EGFR mutants showed increased methylation (p&amp;lt;0.01), while TP53 mutants had lower methylation levels (p&amp;lt;0.05). Conclusions: Our method effectively corrects cfDNA methylation data, even at low tumor concentration, ensuring enhanced methylation signal accuracy and proficient SNV calling capabilities. This integrated approach offers a powerful means for investigating the link between SNVs and DNA methylation, significantly enhancing the accuracy of cancer detection using cfDNA, especially for early-stage tumors. This paves the way for a deeper understanding of the molecular mechanisms driving cancer initiation and progression. Citation Format: Rui Wang, Emily G. Hamilton, Diego Almanza, Angela Hui, Maximilian Diehn, Ash A. Alizadeh. Improved cfDNA methylation profiling through correction of misrepaired jagged-ends [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 960.

Acoustic Path Filtering for Improved Multimode Total Focusing Method Inspection

Total focusing method (TFM) is an ultrasonic testing (UT) technique that provides nondestructive testing (NDT) inspectors with new imaging modes, enabling more accurate detection, sizing, and representation of challenging defects. While TFM may offer convenient, nearly true-to-geometry imagery as the inspection result, it is often detrimentally affected by mode conversion artifacts. Current standards, such as ASME Section V, place the burden on the inspector to explain the origins of those artifacts, impacting the productivity and reliability of the inspection. A method enabling direct control of the ultrasonic wave propagation modes—that is, transverse wave (T) or longitudinal wave (L)—through each interface of the acoustic path is proposed and evaluated in this paper. This control is achieved after the full matrix capture acquisition by modulating, according to the desired propagation mode, the gain applied on the individual paths within the summation process. This leads to the formation of a Path-Filtered Total Focusing Method image. Empirical results on various use cases show considerable improvement of the signal-to-noise ratio through the almost complete elimination of signals originating from undesired paths.

Stratospheric aerosol radiative forcing simulated by the chemistry climate model EMAC using Aerosol CCI satellite data

Abstract. This paper presents decadal simulations of stratospheric and tropospheric aerosol and its radiative effects by the chemistry general circulation model EMAC constrained with satellite observations in the framework of the ESA Aerosol CCI project such as GOMOS (Global Ozone Monitoring by Occultation of Stars) and (A)ATSR ((Advanced) Along Track Scanning Radiometer) on the ENVISAT (European Environmental Satellite), IASI (Infrared Atmospheric Sounding Interferometer) on MetOp (Meteorological Operational Satellite), and, additionally, OSIRIS (Optical Spectrograph and InfraRed Imaging System). In contrast to most other studies, the extinctions and optical depths from the model are compared to the observations at the original wavelengths of the satellite instruments covering the range from the UV (ultraviolet) to terrestrial IR (infrared). This avoids conversion artifacts and provides additional constraints for model aerosol and interpretation of the observations. MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) SO2 limb measurements are used to identify plumes of more than 200 volcanic eruptions. These three-dimensional SO2 plumes are added to the model SO2 at the eruption times. The interannual variability in aerosol extinction in the lower stratosphere, and of stratospheric aerosol radiative forcing at the tropopause, is dominated by the volcanoes. To explain the seasonal cycle of the GOMOS and OSIRIS observations, desert dust simulated by a new approach and transported to the lowermost stratosphere by the Asian summer monsoon and tropical convection turns out to be essential. This also applies to the radiative heating by aerosol in the lowermost stratosphere. The existence of wet dust aerosol in the lowermost stratosphere is indicated by the patterns of the wavelength dependence of extinction in observations and simulations. Additional comparison with (A)ATSR total aerosol optical depth at different wavelengths and IASI dust optical depth demonstrates that the model is able to represent stratospheric as well as tropospheric aerosol consistently.

Prevention of Amino Acid Conversion in SILAC Experiments with Embryonic Stem Cells

Recent studies using stable isotope labeling with amino acids in culture (SILAC) in quantitative proteomics have made mention of the problematic conversion of isotope-coded arginine to proline in cells. The resulting converted proline peptide divides the heavy peptide ion signal causing inaccuracy when compared with the light peptide ion signal. This is of particular concern as it can effect up to half of all peptides in a proteomic experiment. Strategies to both compensate for and limit the inadvertent conversion have been demonstrated, but none have been shown to prevent it. Additionally, these methods combined with SILAC labeling in general have proven problematic in their large scale application to sensitive cell types including embryonic stem cells (ESCs) from the mouse and human. Here, we show that by providing as little as 200 mg/liter L-proline in SILAC media, the conversion of arginine to proline can be rendered completely undetectable. At the same time, there was no compromise in labeling with isotope-coded arginine, indicating there is no observable back conversion from the proline supplement. As a result, when supplemented with proline, correct interpretation of "light" and "heavy" peptide ratios could be achieved even in the worst cases of conversion. By extending these principles to ESC culture protocols and reagents we were able to routinely SILAC label both mouse and human ESCs in the absence of feeder cells and without compromising the pluripotent phenotype. This study provides the simplest protocol to prevent proline artifacts in SILAC labeling experiments with arginine. Moreover, it presents a robust, feeder cell-free, protocol for performing SILAC experiments on ESCs from both the mouse and the human.

Quantitative Proteomics Reveals Regulation of Dynamic Components within TATA-binding Protein (TBP) Transcription Complexes

Affinity purification in combination with isotope labeling of proteins has proven to be a powerful method to discriminate specific from nonspecific interactors. However, in the standard SILAC (stable isotope labeling by amino acids in cell culture) approach dynamic components may easily be assigned as nonspecific. We compared two affinity purification protocols, which in combination revealed information on the dynamics of protein complexes. We focused on the central component in eukaryotic transcription, the human TATA-binding protein, which is involved in different complexes. All known TATA-binding protein-associated factors (TAFs) were detected as specific interactors. Interestingly one of them, BTAF1, exchanged significantly in cell extracts during the affinity purification. The other TAFs did not display this behavior. Cell cycle synchronization showed that BTAF1 exchange was regulated during mitosis. The combination of the two affinity purification protocols allows a quantitative approach to identify transient components in any protein complex.

An experimental correction for arginine-to-proline conversion artifacts in SILAC-based quantitative proteomics

An experimental correction for arginine-to-proline conversion artifacts in SILAC-based quantitative proteomics

Nucleotide analogs facilitate base conversion with 3' mismatch primers.

We compared the efficiency of PCR amplification using primers containing either a nucleotide analog or a mismatch at the 3' base. To determine the distribution of bases inserted opposite eight different analogs, 3' analog primers were used to amplify four different templates. The products from the reactions with the highest amplification efficiency were sequenced. Analogs allowing efficient amplification followed by insertion of a new base at that position are herein termed 'convertides'. The three convertides with the highest amplification efficiency were used to convert sequences containing C, T, G and A bases into products containing the respective three remaining bases. Nine templates were used to generate conversion products, as well as non-conversion control products with no base change. We compared the ability of natural bases to convert specific sites with and without a preconversion step using nucleotide analog primers. Conversion products were identified by a ligation detection reaction using primers specific for the converted sequence. We found that conversions resulting in transitions were easier to accomplish than transversions and that sequence context influences conversion. Specifically, primer slippage appears to be an important mechanism for producing artifacts via polymerase extension of a 3' base or analog transiently base paired to neighboring bases of the template. Nucleotide analogs could often reduce conversion artifacts and increase the yield of the expected product. While new analogs are needed to reliably achieve transversions, the current set have proven effective for creating transition conversions.