last updated: July 25, 2019

Media notebook

Biocomputing is the informational half of nature. It gets its meaning from collective behavior and not from any one individual part. Capturing this in images is not always easy. Here we hope to combine creative microscopy with a little narration and a lot of imagination to make more visible the amazing potential of computing in liquid media.

Image: Yeast functionalized hair strands. Proprietary conjugation technology and fluorescent microscopy. Credit. XENO

Images in this notebook highlight the organizational aspects of our technology. They are selected experimental and theoretical records wherein the spatial and dynamic relationships are at the forefront.

Parallel processing

Living systems analyze themselves. Multicellular organisms are massive parallel computers. Our technology links up to biological objects generating a coherent intelligible interface. It self-assembles into computational clusters and processes information locally within these clusters. Rather than relying on brute-force screening, the linked biological sample reports on itself.

Computational cluster

The fundamental unit of computing is the computational cluster. Each computational cluster may differ in composition. Categorically, a cluster is a physically connected network made up of genetically engineered yeast, linkers, enzymes, and other chemical modifiers. Clusters are not prefabricated. They self-assemble around objects with specific immunological profiles. By fusing different fluorescent labels to different yeast strains, the makeup of a cluster can in part be made visible.

Image: Bivariate computational clusters. Genetically engineered yeast agglutenated with antigen coated beads by complimentary linkers. Antigen specificity of yeast strains is denoted by cyan and yellow fluorescent labels. Cluster colors equate to antigen types presented by the centermost bead. Credit. XENO

High throughput screening revisited

4B cells at 30K cells/sec = 1.5 days of dedicated time on a state-of-the-art flow cytometer. Even a milliliter of blood takes a long time to analyze by brute-force. Extrapolate this number to include cell signaling, adaptation, or differentiation and quickly brute-force analysis becomes intractable. Analysis of a biological sample that has been assimilated into computational clusters scales much better. Each cluster in a shared reaction volume acts autonomously in probing and recording its own behavior. Flagged events are encoded at the cluster level into easily readable signals. An assimilated microliter of blood is analyzed as quickly as a liter.

Image: An assimilated biological sample in a microtitier plate well. Visible objects correspond to approximately 1000 computational clusters. Individual cells are not visible at this scale. Antigen specificity of yeast strains is denoted by fluorescent labels. Credit. XENO

Physical laws of cluster computing

There is a high incentive for engineering mechanisms that improve multicellular behaviors. Yeast cells exchange information with peptide signals. To complete such an exchange, the signal must first accumulate in its cluster without migrating to other clusters. Hence, active and diffusive transport away from the cluster must be low. Factors such as cell density, medium composition, and stratification of media away from the surface contribute to effective signal exchange. Through proper engineering, the maximum number of clusters in a reaction, and thereby the maximum reaction volume and sample size, may be dramatically increased.

Image: Hydrodynamic simulation of a computaional cluster in a liquid medium. Intercellular communication requires high cluster density. Voids between cells permit active transport away from the cluster which attenuates signal concentration. Credit. University of West Bohemia

General logical operations

Composition of computing units prior to sample assimilation dictates possible cluster programs. Both affirmative and negative computing units can be matched to specific target antigens. Connectivity of cluster units is also specified by matching signals to receptors. Upon correct assembly a cluster initiates logical evaluation of cluster properties. Within physical limits any boolean operator may be evaluated in this manner. Cluster computing represents a generalized version of customary gating commonly used to filter flow cytometry data.

Image: Logical operator AND. Time lapse microscopy of a computing cluster evaluating a target for the presence of two specific antigens. Luminescent output indicates communication between strains. Luminescence from both strains denotes an affirmative readout. Credit. XENO

Analyzing the very small

Computational clusters come in all shapes and sizes. The same mixture of computing elements can be used to assimilate samples down to the molecular level. There is an inherent limitation on the resolution of analysis at each level and the logical programs should reflect this. Computing entities can facilitate multivariate analysis on objects of the same size or larger. Below this threshold the analysis is limited to two features per object and any further information must be statistically inferred. For computing yeast the bifurcation occurs approximately at the size of a red blood cell.

Multivariate detection of bacteria

Surface composition and peptide secretion play a crucial role in bacterial pathogenesis. Cluster assimilation enables deconvolution of features that facilitate bacterial adhesion, invasion and evasion. Cluster morphology is highly dependent on the size, distribution, and number of targeted objects. Bacteria at high copy numbers generate clusters that resemble spherical colonies. As the copy number decreases the size of the colony progressively diminishes down to two yeast cells. Bivariate analysis of bacteria sized objects requires approximately 10 such objects to be statistically reliable.

Image: Computing cluster morphology in assimilation of various bacterial samples. Cyan and yellow fluorescent labels indicated genetically engineered yeast and wild type bacterial strains respectively. Credit. XENO

Finding rare cells

Much of how cells navigate a multicellular organism is dictated by their surfaces. Cells present in different tissues adhere using different molecules and express different receptors. Migrating cells are specially equipped to enter and leave the circulatory system. Stem cells sense growth signals and adapt as they transition from one state to another. Mutations or perturbations that generate wrong combinations of surface features may lead to onset or spread of disease. In presymptomatic stages such problematic cells are very rare. Assimilation of bodily fluid samples into clusters permits routine screening, monitoring, and analysis of rare cells.

Image: Mammalian computing clusters. Clones of an immortalized cell line presenting specific antigens are assimilated into computing clusters. Cluster density is artificially reduced to improve visibility of the target cell. Antigen specific yeast strains are marked by fluorescent labels. Credit. XENO

Biocomputing in the visible world

Computational clusters can scale to cover visible surfaces. At this scale it becomes possible to integrate clusters with other technologies, to build entirely new devices, and to construct entirely new fabrication processes. Assembly of large clusters is demonstrated using the same methods as at the microlevel. Realization of such hybrid devices awaits further application platforms and corresponding innovations in biocomputing.

Image: Assimilated hair strands. Functionalized hair cuticles are conjugated to two different genetically engineered yeast strains using proprietary linker technology. Three strands in view are functionalized separately and then assimilated in a single reaction. Credit. XENO

Biocomputing facility

A biocomputing core can massively expand a facility’s potential to screen rare events, correlated events, and dynamic behavior. With biocomputing sparse data is accessed with same throughput as common data. Biocomputing facilities will enable pre-symptomatic screening, personalized medicine, drug discovery, and basic science in general. This will require a paradigm shift in assay design and an upgrade in associated utilities.

Image: XENO lab 1. Lab focuses on internal research evaluating biocomputing solutions for difficult screenign problems. Credit. XENO