Sesame: Crop, Oil, and Multi-Domain Research
- Sesame is an ancient oilseed crop known for its high oil (50%) and protein (25%) content, and is valued for its antioxidative lignans that enhance food quality and pest resistance.
- Metabolic diversity in sesame, revealed through HPLC–ESI–MS and AFLP markers, underscores the need to integrate chemical profiles with genomic data in breeding and conservation.
- Beyond agriculture, 'Sesame' spans diverse applications including sesame oil in pressure ulcer prevention, agricultural AI datasets, comet lander instruments, and computational models in various tech fields.
Sesame most commonly denotes Sesamum indicum L., an oilseed crop described as one of the most ancient crops and valued because its seeds are used directly as food or pressed for high-quality oil. In the research literature represented here, the term also appears as SESAME, Sesame, and SeSame in a series of unrelated technical names: a nursing intervention based on sesame oil, a sesame-plant vision dataset, the Surface Electric Sounding and Acoustic Monitoring Experiment on Rosetta’s Philae lander, and multiple software or model names in semiconductor physics, cosmology, computer vision, smart-home security, secure accelerators, and drug discovery (Laurentin et al., 2014, Damanik et al., 2020, Muhammad et al., 12 Jan 2026, Krüger et al., 2015, Gaury et al., 2018, Mauland et al., 2023).
1. Botanical crop and seed chemistry
In crop biology, sesame is treated as a chemically rich oilseed species whose commercial importance is not exhausted by yield alone. The seeds are described as containing about 50% oil and 25% protein, and as being rich in secondary metabolites, especially the lignans sesamin, sesamol, sesamolin, and sesaminol. These lignans are reported to have antioxidative properties, to contribute to health-promoting effects of sesame food products, and potentially to participate in defense against insects and microbial pathogens (Laurentin et al., 2014).
The same source emphasizes that sesame is a useful example of a crop in which chemical phenotype is agriculturally consequential. Traits such as taste, color, aroma, oxidative stability, and pest resistance may depend more directly on metabolite composition than on neutral DNA variation. This is important for breeding and conservation because the crop’s economically relevant phenotype is partly encoded at the level of metabolism rather than only at the level of genome polymorphism (Laurentin et al., 2014).
2. Diversity, metabolomics, and breeding relevance
A detailed diversity study compared genomic diversity and metabolic diversity across 10 sesame accessions chosen to represent much of the known genome diversity of sesame grown in India, Western Asia, Sudan, and Venezuela. Genomic variation was measured by AFLP using eight primer-pair combinations, yielding 381 AFLP markers, of which 95% were polymorphic and 88 bands (23%) were unique. Metabolic variation was measured by untargeted HPLC–ESI–ion trap MS on ethanolic seed extracts, retaining 88 dominant metabolic signals, including 34 signals common to all accessions (Laurentin et al., 2014).
The central result was that AFLP-based genomic similarity was not significantly correlated with metabolic similarity. The reported correlations were for AFLP versus the metabolite simple-matching matrix and for AFLP versus the correlation-based metabolite matrix. By contrast, two internal codings of metabolic diversity were moderately consistent, with a correlation of 0.63 (). The authors interpret the divergence between genome and metabolome patterns as evidence that selection played a significant role in the evolution of metabolic diversity in sesame, and conclude that AFLP fingerprints should be complemented with metabolic profiles when selecting material for breeding and conservation (Laurentin et al., 2014).
This result also addresses a common simplification in crop diversity studies: neutral markers do not necessarily recover the variation most relevant for chemistry-dependent agronomic traits. A plausible implication is that sesame is a model case for long-domesticated crops in which metabolite profiles carry information not recoverable from neutral DNA markers alone.
3. Sesame oil in pressure-ulcer prevention
In clinical nursing research, sesame appears as sesame oil, evaluated as a topical intervention for the prevention of pressure ulcers in hospitalized patients confined to bed rest. Pressure ulcers are framed as a major nursing problem because prolonged pressure reduces tissue blood flow and can lead to skin breakdown, pain, infection risk, prolonged hospitalization, and increased nursing burden. The stated rationale for sesame oil is its proposed anti-inflammatory, antioxidant, and analgesic properties and its content of fatty acids, vitamins, minerals, and lignans such as sesamin, episesamin, and sesamolin (Damanik et al., 2020).
The study was a randomized controlled trial with a parallel design without matching, conducted at Aji Batara Dewa Sakti Hospital, Samboja, from July to August 2020. The sample comprised 40 inpatients with bed rest, randomized into an intervention group of 20 and a control group of 20. Eligibility required inpatient bed rest, willingness to participate, risk assessment by the NPUAP scale, use of the standard hospital bed and mattress, a negative allergy test, and absence of special treatment for pressure ulcers. Both groups received standard preventive care: repositioning every 2 hours, lateral tilting left and right and supine positioning, pillow support so the ischium and sacrum were lifted 30°, and bathing twice daily with a washcloth and soap. The intervention group additionally received a light massage/backrub with sesame oil over the back from the scapula to the ischium and over the heel to malleolus area, performed 2 times after bathing, with observation for 3 days (Damanik et al., 2020).
The reported between-group difference in pressure-ulcer incidence was statistically significant by Chi-square, with . The paper also reported with 95% CI: , interpreted by the authors as indicating that patients who did not receive sesame oil together with positioning care had about 9.333 times the chance of developing pressure ulcers compared with those who did receive the sesame-oil intervention. The discussion also states that 80% of respondents experienced a decrease in the degree of pressure sores after sesame-oil administration. The authors therefore conclude that skin care with sesame oil can prevent pressure ulcers and recommend it as a nursing intervention for bedridden hospitalized patients (Damanik et al., 2020).
The observation period was only 3 days and the sample size was 40, so the result is best read as evidence from a small RCT rather than as a definitive clinical standard. Even so, the care bundle tested in the study makes the intervention operationally precise: routine positioning, hygiene, and a sesame-oil backrub twice daily after bathing.
4. Sesame in agricultural computer vision
In agricultural AI, “sesame” names the target crop of the Sesame Plant Segmentation Dataset, an open-source image dataset created for YOLO-compatible segmentation of sesame plants in Nigerian field conditions. The dataset contains 206 training images, 43 validation images, and 43 test images, for a total of 293 annotated images. Images were collected from farms in Jirdede, Daura Local Government Area, Katsina State, Nigeria, using a high resolution mobile camera, specified in the details as an iPhone 11 camera with wide and ultra-wide 12-megapixel lenses. The plants were captured at early growth stages (45–85 days) under natural daylight during the rainy season, from horizontal, top, and side views, and about 95% of the plants were healthy and unaffected by disease (Muhammad et al., 12 Jan 2026).
The distinguishing technical feature of the dataset is pixel-level instance segmentation rather than bounding-box annotation. Annotation was performed with the Segment Anything Model version 2 (SAM-2) on the Roboflow platform, with farmer supervision. This is significant because sesame plants can exhibit overlapping leaves, dense clusters, and irregular shapes, for which bounding boxes are coarse. The train/valid/test split follows a standard 70% / 15% / 15% organization, and the data are arranged in YOLO segmentation directories with paired images and labels (Muhammad et al., 12 Jan 2026).
For validation, the authors trained Ultralytics YOLOv8 for 100 epochs with batch size 16, image resolution 640×640, and optimizer “auto”. For bounding-box detection, the model achieved Recall 79%, Precision 79%, mAP@50 84%, and mAP@50-95 58%. For segmentation, it achieved Recall 82%, Precision 77%, mAP@50 84%, and mAP@50-95 52%. The paper presents the dataset as a novel contribution and as the first localized dataset focused on a Nigerian sesame crop and annotated specifically for real-time segmentation tasks, with intended applications in plant monitoring, yield estimation, smart spraying systems, weed detection, agricultural research, and automation (Muhammad et al., 12 Jan 2026).
5. SESAME on Rosetta’s Philae lander
In planetary science, SESAME stands for Surface Electric Sounding and Acoustic Monitoring Experiment, one of the scientific payloads on the Rosetta lander Philae. SESAME consisted of three instruments: CASSE for acoustic characterization of the comet subsurface, DIM for dust impact detection, and PP for permittivity and electrical sounding. Within this suite, DIM—the Dust Impact Monitor—was designed to detect impacts of sub-millimeter and millimeter-sized dust and ice particles emitted from comet 67P/Churyumov-Gerasimenko (Krüger et al., 2015).
DIM employed piezoelectric PZT sensors mounted on a cube about 7 cm on a side, with three active faces aligned with Philae’s +X, +Y, and +Z directions and a total active area of about 70 cm². The sensor measured the peak voltage and impact duration , from which particle size and speed could be constrained under Hertzian contact assumptions. DIM was operated in three mission phases: before separation from Rosetta, during descent to the nominal landing site Agilkia, and at the final landing site Abydos. No dust particles were detected before separation; during descent, DIM detected one approximately millimeter-sized particle on 12 November 2014 at 14:43:47 UTC on the +Y sensor at a distance of 5.0 km from the nucleus barycenter and an altitude of about 2.4 km above the surface; at Abydos, no dust impacts were detected (Krüger et al., 2015).
The single descent event is described as the closest ever in situ dust detection at a cometary nucleus by a dedicated dust detector. Its measured signal was , corresponding to and 0. Comparison with laboratory calibration experiments suggested a porous particle compatible with a bulk density of approximately 1, with radius 2 mm and impact speed around 3 under the nominal aerogel-based interpretation. The authors argue that such a particle could have been lifted from the comet surface by sublimating water ice (Krüger et al., 2015).
A later analysis focused on flux upper limits rather than on single-particle reconstruction. With the relation 4 between impact rate, geometric factor, and directional flux, and with Poisson counting for intervals with zero detections, the authors derived geometric factors of 34.0, 47.9, and 51.0 cm5 sr for the X, Y, and Z sides in the full lander configuration. The upper limit of the particle flux in DIM’s measurement range was of order 6 to 7 during descent and 8 at Abydos on 13–14 November 2014. Assuming particle speeds below escape velocity, the corresponding upper boundary for particle volume density was constrained to approximately 9 to 0. Simulations with the GIPSI tool predicted only about 0.005 to 0.5 expected counts over the usable descent interval, which the authors considered compatible with DIM’s essentially null result (Hirn et al., 2016).
6. “Sesame” as a recurrent technical name in computation, security, and molecular design
Beyond the crop and the Philae payload, Sesame/SESAME/SeSame is reused as a project name across several technical domains. The following instances are explicitly represented in the cited literature.
| Name | Domain | Stated function |
|---|---|---|
| Sesame (Gaury et al., 2018) | Semiconductor modeling | Numerical computation of classical semiconductor equations in 1D and 2D |
| Sesame (Woszczyk et al., 2021) | Voice-service security | Fine-grained access control for smart-home voice commands |
| SESAME (Ntavelis et al., 2020) | Image synthesis | Semantic editing of scenes by adding, manipulating, or erasing objects |
| SESAME (Banerjee et al., 2020) | Secure accelerators | Software-defined enclaves for multi-tenant inference accelerators |
| Sesame (Mauland et al., 2023) | Cosmology | Emulator pipeline for beyond-1CDM matter power spectra |
| SeSame (O et al., 2024) | LiDAR perception | 3D object detection with point-wise semantics |
| Sesame (Miñán et al., 21 Aug 2025) | Structure-based drug discovery | Generative model for pocket-opening protein conformations |
| Sesame (Yatsenko et al., 22 Jun 2026) | Molecular generation | Structure-aware diffusion model using spatial density-map conditioning |
In semiconductor physics, the software package Sesame solves the steady-state drift-diffusion-Poisson system in 1D and 2D, supports grain boundaries and sample surfaces, and uses finite differences, the Scharfetter–Gummel scheme, and a Newton–Raphson solver with an analytically computed Jacobian. It is written in Python, distributed under the BSD license, and validated against SCAPS, Sentaurus, and COMSOL, with reported maximum relative differences of about 0.2% versus Sentaurus and 2% versus COMSOL in a 1D CdS/CdTe benchmark, and about 0.5% versus Sentaurus and 0.7% versus COMSOL in a 2D grain-boundary benchmark. In cosmology, Sesame denotes a pipeline that emulates the boost 2, enabling non-linear power-spectrum emulators for models beyond 3CDM without supercomputer-scale simulation suites. The demonstration case used approximately 3000 CPU hours, 550 Latin-hypercube samples, and reported useful reach to 4, with the linear boost emulator below 1% error across the tested scales and redshifts and the non-linear cases mostly below 1–2.5% depending on redshift and screening treatment. In computer architecture, SESAME names a confidential-computing design for accelerators based on software-defined enclaves, adding private queues, traffic shaping, scratchpad partitioning, zeroization, and constant-time instructions; the reported code-size increase is 3–7%, and performance overhead for specific defenses ranges from 3.96% to 34.87% across threat models (Gaury et al., 2018, Mauland et al., 2023, Banerjee et al., 2020).
In AI and security, the name is equally heterogeneous. The voice-security framework Sesame combines Automatic Speech Recognition, Natural Language Understanding, and a Policy module to enforce fine-grained authorization over Alexa and Google Home commands. Implemented on Android with local DeepSpeech ASR and BERT/MobileBERT-based NLU, it reports 362 ms end-to-end inference and a <25 MB NLU model in its lightweight configuration. The image-editing system SESAME is a conditional GAN for adding, manipulating, or erasing semantic concepts; its central architectural claim is a two-stream discriminator that processes image and semantics separately and uses the latter to modulate the former. On Cityscapes object addition, the full SESAME configuration with BBox semantics reports SSIM 0.410, accuracy 86.0, mIoU 65.3, and FID 11.03; on object removal, it reports SSIM 0.797, accuracy 85.0%, mIoU 67.6%, and FID 7.43. In LiDAR perception, SeSame augments each point with one-hot semantic labels predicted by Cylinder3D, making the feature vector 5. On KITTI, SeSame + Point reports car AP6 of 85.25 / 76.83 / 71.60 for easy/moderate/hard, compared with 85.94 / 75.76 / 68.32 for PointRCNN, while SeSame + voxel improves hard car AP7 from 66.20 to 70.53 over SECOND (Woszczyk et al., 2021, Ntavelis et al., 2020, O et al., 2024).
In drug-discovery settings, two later systems reuse the name for structure generation problems. One Sesame is a flow-matching model that predicts apo-to-holo-like conformational changes in protein pockets; on D3PM-Large, it reports median RMSD 2.87 Å and median 8, outperforming SBAlign at 3.67 Å and 1.30 Å respectively, and on PDBBind-MD it reports median RMSD 0.18 Å. The other expands the name as Spatial Evoformer for a Structure-Aware Molecular Engine and uses 0.5 Å density maps on a 9 voxel cube to condition a diffusion model for de novo generation and fragment-conditioned lead optimization. It reports 94.8% fragment retention, and after simple post-processing the validity rises to 92.4% for protein + fragment lead optimization and 88.7% for protein-only de novo generation (Miñán et al., 21 Aug 2025, Yatsenko et al., 22 Jun 2026).
Taken together, these usages show that “Sesame” is not a single research object but a polysemous label. In one literature it refers to an ancient oilseed crop with chemically important lignans and selection-shaped metabolic diversity; in another it denotes sesame oil as a nursing intervention; elsewhere it appears as a dataset name, a comet-lander instrument suite, and a recurring acronym or project name for tools that measure, model, secure, segment, generate, or authorize complex systems.