Molecular tomography

Molecular tomography is Molecular Reality Corporation's name for a proposed way of identifying a single molecule: measure it in a nanopore many times, under different conditions, and combine the measurements, in the way a CT scanner combines X-ray projections taken from many angles. Company materials of July 2026 put it in four sentences: "A CT scanner cannot tell what's inside you from one X-ray. It takes many projections from many angles and reconstructs the interior. We intend to do the same to a molecule, in current instead of X-rays, one clean event at a time."^[1] The company has not reported a reconstruction.

A molecule passing through a nanopore lowers the ionic current for as long as it is inside, and the depth and duration of the dip are most of what the instrument records. The company's critique of its own machine-learning plans, published under Keltar's name, states the limitation as a matter of geometry: "A cylinder casts a circular shadow from one angle and a square from another. You cannot recover a 3-D object from one shadow. Ionic current is a 1-D shadow of a 3-D molecule." Its remedy is the tomographer's: "you must cast more shadows from more angles."^[2] The same page calls identifying any molecule "an inverse problem", in which many different molecules can produce the same trace.

It grew from an older idea. The company's December 2022 overview described the blur that Brownian motion gives to a single measurement as "something like how faces are blurry in old photographs", and the cure as to "take a hundred photographs, and mathematically reconstruct the faces." It identified the nanopore equivalent as passing a molecule back and forth through the sensing zone, a method "literally called 'ping-pong' by some workers".^[3] Repeated passes under fixed conditions improve the precision of one number. In 2015 a multipass Coulter counter reached a precision "statistically equivalent to resolving particle radii differing by merely 31 pm", which a company post noted is "smaller than a single hydrogen atom".^[4][5] Molecular tomography asks for something different: passes under changed conditions, each of which measures a different property.

In 1917 the mathematician Johann Radon showed that a function on a plane can be recovered from its integrals along all the straight lines through it. An X-ray beam measures one such integral, the total attenuation along its path, and a row of parallel beams at one angle gives a projection. Allan Cormack worked out the reconstruction for X-ray absorption in papers of 1963 and 1964, and Godfrey Hounsfield, an engineer at EMI, built the first clinical scanner, which imaged its first patient in London in October 1971. The two shared the Nobel Prize in Physiology or Medicine in 1979.

A scanner gets its information from the variety of angles. Each projection, transformed into frequencies, fills one line through the center of the object's two-dimensional frequency map, and enough angles fill the map, which is then inverted. One projection gives one line, however many times it is repeated.

In a pore the angles are conditions the instrument can change between passes: the applied voltage, the salt solution and, in the company's designs, the shape of the pore itself. Sliding occlusion changes the size of the opening by moving one membrane over another, and with three or more membranes the length of the sensing zone as well, so the same molecule could be measured through a series of different openings. The company's page "Molecular Self-Play" animates the cycle under the caption "TRAP · SENSE · RELEASE · REPEAT" and describes a pore that "traps a single construct, ping-pongs the analyte through the sensing zone to tomographically sub-sample different conformations, then releases it".^[6] Choosing the next condition on the basis of the last measurement is the subject of active sensing and ping-pong sensing.

The analogy has limits that CT does not face. A scanner sets each angle and knows it, and the patient holds still. A molecule enters a pore at an orientation nobody chose, may change shape on the way through and leaves within microseconds to milliseconds. Reconstruction from views at unknown orientations is the problem that single-particle cryo-electron microscopy solves with images of many copies of a molecule; the company proposes to work with one copy at a time.

Molecular tomography belongs to the company's proposals for machine learning, with The Nanopore World Model and RLMF, which its guide introduces with: "Both are proposals, not products, and I want that said first."^[7] The pores on the company's live stream in 2026 were larger than a nanometer, and apertures below 15 nanometers remained a target.

• Ideas and influences of Molecular Reality Corporation
• Speeding bullet problem
• Translocation (nanopore)