Beam
Beam Therapeutics Inc. is a biotechnology company that develops precision genetic medicines through its proprietary base editing platform, a CRISPR-derived technology enabling targeted single-nucleotide changes in DNA without inducing double-strand breaks.[1][2]Founded in 2017 in Cambridge, Massachusetts, by gene editing pioneers including David R. Liu, J. Keith Joung, and Feng Zhang, the company seeks to address serious diseases such as sickle cell disease, glycogen storage disease Ia, and alpha-1 antitrypsin deficiency by creating a new class of therapies that offer greater precision and potentially reduced off-target effects compared to traditional gene editing methods.[3][4][5]
Beam Therapeutics went public on the NASDAQ exchange under the ticker BEAM in early 2020, raising significant capital to advance its pipeline, which includes investigational programs in liver, hematologic, and immunological disorders, reflecting its commitment to translating foundational scientific discoveries into clinical applications.[6][7]
The firm's base editing innovations, stemming from academic breakthroughs by its founders, position it as a leader in next-generation genetic engineering, though like many biotechs, its therapeutic candidates remain in preclinical and early clinical stages amid the inherent risks of drug development.[1][8]
Structural and mechanical applications
In construction and engineering
In structural engineering, a beam is a horizontal structural element designed primarily to resist bending and shear forces induced by transverse loads, transferring these forces to vertical supports such as columns or walls. Beams support vertical loads from floors, roofs, or other elements, undergoing deflection, bending moments, and shear stresses in the process.[9][10] Beams are classified by their support conditions and loading configurations, which determine internal force distributions and deflection profiles. Common types include:- Simply supported beams, which rest on supports at both ends, allowing rotation but not vertical displacement; these experience maximum bending moments at midspan under uniform loading.[11][12]
- Cantilever beams, fixed rigidly at one end and free at the other, used for projections like balconies; they develop maximum moments at the fixed support.[13][11]
- Continuous beams, spanning multiple supports, which distribute loads more evenly but introduce negative moments over intermediate supports.[14][15]
- Fixed beams, restrained against rotation and displacement at both ends, minimizing deflection but increasing end moments.[11]
- Overhanging beams, extending beyond supports, combining simply supported and cantilever behaviors.[15][11]
In naval architecture
In naval architecture, the beam of a vessel denotes its maximum width, measured perpendicular to the centerline at the widest point of the hull.[25] This dimension is typically specified as the moulded beam, taken from the inner surface of the shell plating on one side to the corresponding point on the opposite side, excluding any protruding fittings or fenders.[25] The beam at the waterline (BWL), often amidships, represents the breadth at the design or summer load waterline, influencing hydrostatic properties such as the waterplane area and moment of inertia.[25][26] A wider beam enhances transverse stability by increasing the vessel's righting moment against heeling forces, primarily through a larger transverse metacentric height (BM), calculated as the second moment of the waterplane area divided by the displaced volume.[25][27] For instance, catamarans and multihull designs exploit greater beam relative to length (low length-to-beam ratio) to achieve inherent stability without relying heavily on ballast or deep keels.[28] Conversely, excessive beam can elevate frictional and wave-making resistance, reducing speed and fuel efficiency, particularly in slender hull forms optimized for hydrodynamic performance.[25][29] Beam also directly impacts cargo capacity and structural design; broader hulls accommodate larger deck areas and internal volumes for holds or compartments, but necessitate reinforced framing to counter bending stresses amidships.[30][31] In regulatory contexts, the extreme beam (including overhangs) determines lock or canal compatibility, while registered beam—adjusted for tax or tonnage calculations—may exclude certain protrusions.[25] Optimal beam selection balances these factors against length and depth, with empirical data from model testing and computational fluid dynamics guiding ratios like 6:1 to 8:1 for conventional merchant ships to minimize resistance while ensuring stability margins.[28][29]Physical beams of energy or particles
Electromagnetic beams
Electromagnetic beams are collimated streams of electromagnetic radiation, characterized by a confined transverse intensity profile and directed propagation, enabling low divergence and focused energy delivery compared to isotropic or plane wave emissions.[32] These beams arise from the wave nature of electromagnetic fields, where oscillating electric and magnetic components perpendicular to the direction of travel maintain coherence over distance when generated by sources like lasers or phased-array antennas.[33] In free space, their behavior follows Maxwell's equations, with diffraction limiting perfect collimation; the fundamental solution for paraxial propagation is the Gaussian beam mode.[32] The Gaussian beam represents the ideal transverse electromagnetic (TEM00) mode, with electric field amplitude E(r,z) \propto \exp\left( -r^2 / w(z)^2 \right) \exp\left( i k z + i \phi(z) - i k r^2 / (2 R(z)) \right), where r is radial distance, z axial, w(z) the beam radius, R(z) radius of curvature, and \phi(z) Gouy phase.[32] The beam waist w_0 at z=0 defines the minimum spot size, with Rayleigh range z_R = \pi w_0^2 / \lambda marking the distance of near-constant width before divergence \theta \approx \lambda / (\pi w_0) dominates.[32] Higher-order modes (e.g., Hermite-Gaussian) exhibit structured profiles but increased divergence, reducing beam quality quantified by the M^2 factor, where M^2 = 1 for ideal Gaussian.[32] Generation spans the spectrum: low-frequency beams (radio, microwave) via aperture antennas achieve directionality through phased arrays, with beamwidth \Delta \theta \approx \lambda / D for aperture diameter D; optical beams from lasers exploit stimulated emission for coherence.[34] The first laser beam, producing a pulsed ruby output at 694 nm, was achieved by Theodore H. Maiman on May 16, 1960, using a helical flashlamp-pumped synthetic ruby crystal, marking the advent of controllable high-intensity optical beams. Continuous-wave gas lasers followed in December 1960, enabling stable beams for precision uses.[35] Applications leverage directionality and coherence: optical beams enable laser machining with kerf widths below 0.1 mm, ablation rates up to 10 mm³/s in metals, and non-thermal precision in surgery (e.g., retinal procedures since the 1960s); microwave beams support radar ranging with resolutions to meters and satellite communications achieving gigabit data rates; X-ray beams from synchrotrons facilitate atomic-scale imaging in materials science.[34] Free-space optical links using Gaussian beams transmit terabits per second over kilometers with bit error rates below 10-9, outperforming radio in bandwidth but sensitive to atmospheric scintillation.[32] Emerging uses include directed-energy weapons, where megawatt-class beams propagate kilowatts over kilometers, limited by thermal blooming.[32]Particle beams
Particle beams consist of streams of subatomic particles, such as electrons, protons, or ions, accelerated to high velocities using electric and magnetic fields.[36] These beams can be charged, allowing manipulation via electromagnetic forces but prone to divergence from mutual repulsion, or neutral, formed by neutralizing charged particles to reduce spreading while complicating steering. Acceleration occurs in particle accelerators, where particles gain energy through repeated interactions with oscillating electric fields, reaching relativistic speeds in facilities like cyclotrons or synchrotrons.[37] In scientific research, particle beams enable high-energy collisions to probe fundamental physics. The Large Hadron Collider (LHC) at CERN accelerates proton beams to 6.5 teraelectronvolts (TeV) per beam, colliding them head-on to recreate early universe conditions and discover particles like the Higgs boson in 2012.[38] Beam dynamics in colliders manage intense power densities, with luminosities exceeding 10^34 interactions per square centimeter per second to maximize event rates.[39] Medical applications leverage the Bragg peak, where charged particle beams deposit maximum energy at a precise depth before stopping abruptly, minimizing damage to surrounding tissue. Proton therapy, using beams of hydrogen ions accelerated to energies around 70-250 MeV, treats over 100,000 patients annually worldwide for cancers like prostate and brain tumors, with facilities such as Mayo Clinic's proton beam program delivering targeted doses since the 1990s.[40] Carbon ion beams extend this to heavier particles for radioresistant tumors, achieving superior local control rates in trials, such as 81% for chordomas versus 58% with conventional radiotherapy.[41] Industrial uses include electron beams for materials processing, such as polymer cross-linking to enhance strength or welding thick metals without filler materials, operating at currents up to kiloamperes and energies of 1-10 MeV.[42] In defense research, neutral particle beam concepts were explored under the U.S. Strategic Defense Initiative in the 1980s for space-based interception of missiles, accelerating hydrogen atoms to gigaelectronvolt energies, though technical challenges like beam neutralization efficiency limited deployment.[43] No operational particle beam weapons exist as of 2025, with efforts constrained by atmospheric scattering and power requirements.[44]Computing and software frameworks
Apache Beam
Apache Beam is an open-source unified programming model and software development kit (SDK) designed for defining and executing both batch and streaming data processing pipelines.[45] It enables developers to write portable code that handles large-scale data workflows, including extract-transform-load (ETL) operations, using a consistent abstraction across diverse execution environments.[46] The model abstracts away low-level details of distributed computing, focusing instead on high-level pipeline definitions composed of data collections and transformations. The project originated from Google's internal data processing technologies, including FlumeJava and the Cloud Dataflow service, which were formalized in the 2015 Dataflow model research paper.[47] In February 2016, Google donated the Cloud Dataflow SDKs and related components to the Apache Software Foundation as an incubating project, aiming to standardize batch and streaming processing beyond proprietary systems.[47] The initial release, version 0.1.0-incubating, occurred on June 15, 2016, marking the project's entry into open-source development.[48] By January 10, 2017, Apache Beam achieved top-level project status within the Apache Foundation, signifying broad community governance and maturity.[49] At its core, Apache Beam's programming model revolves around three primary abstractions: pipelines, which represent the overall data processing workflow; PCollections, immutable representations of distributed datasets that can be bounded (batch) or unbounded (streaming); and PTransforms, functions that apply operations such as mapping, filtering, grouping, or joining to PCollections.[46] This model supports windowing for time-based grouping in streaming data, watermarking for handling late arrivals, and triggers for output emission control, ensuring deterministic results even in unbounded scenarios. Language-specific SDKs exist for Java, Python, Go, and Scala, allowing pipeline authoring in familiar paradigms while maintaining portability.[45] A key strength of Apache Beam is its runner portability, enabling the same pipeline code to execute on multiple backends without modification, including the reference DirectRunner for local testing, Apache Spark, Apache Flink, Hazelcast Jet, and cloud-native services like Google Cloud Dataflow.[45] This decoupling of pipeline logic from execution engines reduces vendor lock-in and facilitates hybrid batch-streaming applications. I/O connectors integrate with sources such as Apache Kafka, Google BigQuery, and Amazon Kinesis, supporting diverse data ingestion and egress. Adoption spans enterprises like LinkedIn, which processes trillions of events daily using Beam for unified analytics, highlighting its scalability for production workloads.[50]BEAM virtual machine
The BEAM virtual machine is the bytecode interpreter and runtime component of the Erlang Runtime System (ERTS), responsible for executing compiled Erlang and Elixir code in a concurrent, distributed environment.[51] It compiles source code into platform-independent bytecode stored in .beam files, which are then loaded and interpreted to support lightweight processes, message passing, and fault isolation without shared mutable state.[52] BEAM operates as a register-based machine, distinct from stack-based VMs like the JVM, enabling efficient handling of Erlang's actor-like concurrency model where processes communicate asynchronously via messages rather than threads sharing memory.[51][53] Originally termed Bogdan's Erlang Abstract Machine after developer Bogumil "Bogdan" Hausman, BEAM succeeded earlier Erlang VMs such as JAM (Joe's Abstract Machine) introduced in 1989, which offered a 70-fold speedup over the initial Prolog-based implementation but lacked BEAM's threaded code efficiency.[54] BEAM was developed as a hybrid native/threaded machine for superior performance, stabilizing in OTP R1B around 1996 for production use at Ericsson.[54] Key evolutions include the modern .beam file format in OTP R5B (1998), introduction of Core Erlang as an intermediate language in later releases for better optimizations, and reduction of the instruction set from over 300 to 159 active instructions by OTP 20 (2017), incorporating features like constant folding and pattern matching compilation.[54] These changes prioritized execution speed, code loading efficiency, and compatibility across embedded devices to multicore servers.[52] In architecture, BEAM employs X registers for temporary values passed between functions and Y registers for locals bound to stack frames, with instructions operating directly on these registers to avoid stack manipulation overhead.[51] Code execution uses direct threaded interpretation, where instructions are pointers to native code, supporting optimizations like peephole instruction combining (e.g., mergingmove and return into move_return) and specialization for register pairs.[52] Garbage collection is per-process and generational-copying, triggered by dedicated instructions like gc_bif that preserve live X registers while minimizing pauses to milliseconds, isolating failures to individual processes without system-wide stops.[51][53]
Concurrency in BEAM relies on ERTS-managed lightweight processes—green threads consuming minimal memory (around 2-3 KB each)—allowing millions to run concurrently via preemptive scheduling with reduction counts (typically 4000 per timeslice) that yield control at function calls or message receives.[53][52] Multiple schedulers, one per core, balance runnable processes across per-processor queues, enabling true parallelism on multicore hardware while processes remain isolated to prevent cascading failures.[52] Distribution extends this model across nodes via transparent message passing, with hot code loading permitting module updates without downtime by resolving remote calls to the latest version.[53][52] These traits underpin BEAM's suitability for high-availability systems, such as telecommunications and real-time applications, where fault-tolerance is enforced through supervisor hierarchies rather than low-level error handling.[53]