McCue, C. | Search bioRxiv

bioRxiv2024

A molecular representation system with a common reference frame for natural products pathway discovery and structural diversity tasks.

Researchers have uncovered hundreds of thousands of natural products, many of which contribute to medicine, materials, and agriculture. However, missing knowledge of the biosynthetic pathways to these products hinders their expanded use. Nucleotide sequencing is key in pathway elucidation efforts, and analyses of natural products molecular structures, though seldom discussed explicitly, also play an important role by suggesting hypothetical pathways for testing. Structural analyses are also important in drug discovery, where many molecular representation systems - methods of representing molecular structures in a computer-friendly format - have been developed. Unfortunately, pathway elucidation investigations seldom use these representation systems. This gap is likely because those systems are primarily built to document molecular connectivity and topology, rather than the absolute positions of bonds and atoms in a common reference frame, the latter of which enables chemical structures to be connected with potential underlying biosynthetic steps. Here, we present a unique molecular representation system built around a common reference frame. We tested this system using triterpenoid structures as a case study and explored the systems applications in biosynthesis and structural diversity tasks. The common reference frame system can identify structural regions of high or low variability on the scale of atoms and bonds and enable hierarchical clustering that is closely connected to underlying biosynthesis. Combined with phylogenetic distribution information, the system illuminates distinct sources of structural variability, such as different enzyme families operating in the same pathway. These characteristics outline the potential of common reference frame molecular representation systems to support large-scale pathway elucidation efforts. Significance StatementStudying natural products and their biosynthetic pathways aids in identifying, characterizing, and developing new therapeutics, materials, and biotechnologies. Analyzing chemical structures is key to understanding biosynthesis and such analyses enhance pathway elucidation efforts, but few molecular representation systems have been designed with biosynthesis in mind. This study developed a new molecular representation system using a common reference frame, identifying corresponding atoms and bonds across many chemical structures. This system revealed hotspots and dimensions of variation in chemical structures, distinct overall structural groups, and parallels between molecules structural features and underlying biosynthesis. More widespread use of common reference frame molecular representation systems could hasten pathway elucidation efforts.

biochemistry↗

bioRxiv2024

Baikal: Unpaired Denoising of Fluorescence Microscopy Images using Diffusion Models

Fluorescence microscopy is an indispensable tool for biological discovery but image quality is constrained by desired spatial and temporal resolution, sample sensitivity, and other factors. Computational denoising methods can bypass imaging constraints and improve signal-tonoise ratio in images. However, current state of the art methods are commonly trained in a supervised manner, requiring paired noisy and clean images, limiting their application across diverse datasets. An alternative class of denoising models can be trained in a self-supervised manner, assuming independent noise across samples but are unable to generalize from available unpaired clean images. A method that can be trained without paired data and can use information from available unpaired highquality images would address both weaknesses. Here, we present Baikal, a first attempt to formulate such a framework using Denoising Diffusion Probabilistic Models (DDPM) for fluorescence microscopy images. We first train a DDPM backbone in an unconditional manner to learn generative priors over complex morphologies in microscopy images. We then apply various conditioning strategies to sample from the trained model and propose an optimal strategy to denoise the desired image. Extensive quantitative comparisons demonstrate better performance of Baikal over state of the art self-supervised methods across multiple datasets. We highlight the advantage of generative priors learnt by DDPMs in denoising complex Flywing morphologies where other methods fail. Overall, our DDPM based denoising framework presents a new class of denoising methods for fluorescence microscopy datasets that achieve good performance without collection of paired high-quality images. Github repo: https://github.com/scelesticsiva/denoising/tree/main

bioinformatics↗