# Eman Ahmed -- PhD Candidate in Biomedical Engineering Source: https://emanahmed.org/ · Licence: CC BY 4.0 · Citation metrics as of September 2026 ## Profile ### Eman Ahmed -- PhD Candidate in Biomedical Engineering https://emanahmed.org/ Eman Ahmed is a PhD Candidate in Biomedical Engineering in the Gormley Lab at the Department of Biomedical Engineering, Rutgers, The State University of New Jersey, in Piscataway, NJ. Advisor: Adam J. Gormley. Research: high-throughput polymer chemistry, laboratory automation and machine learning for biomaterials discovery and protein stabilization. As of September 2026: 3 publications, 43 citations, h-index 3. Contact: eman.ahmed@rutgers.edu. ## Research areas ### Polymer-stabilized enzymes in organic solvents https://emanahmed.org/research.html#protein-stabilization Doctoral project, Gormley Lab, 2022-2026 Enzymes are extraordinary catalysts in water and frequently useless outside it. Move one into a water-miscible organic solvent, which is often where the interesting synthetic chemistry happens, and it tends to unfold, aggregate and drop out of solution. Random copolymers can act as synthetic chaperones, wrapping a protein in a shell whose chemistry can be tuned monomer by monomer. The difficulty is that the design space is enormous and the structure-function relationship is not obvious from first principles. My work builds plate-based, automated assays that measure solubility and retained activity across large copolymer libraries, so the search can be run empirically at scale rather than one rational design at a time. ### Machine learning for biomaterial structure-function mapping https://emanahmed.org/research.html#machine-learning First-author review, Tissue Engineering Part A, 2024 High-throughput experiments produce the kind of data that models need: many conditions, consistent measurement, and, importantly, retained failures. The question is what to do with it. My review in Tissue Engineering Part A surveys how machine learning is being used to identify the physicochemical cues that govern biomaterial performance, across tissue engineering, gene delivery, drug delivery, protein stabilization and antifouling surfaces, along with the data-mining approaches that let published results stand in for experiments that have not been run yet. ### Automated polymer synthesis and analysis https://emanahmed.org/research.html#automation photo-ATRP platform and SAXS Assistant Oxygen-tolerant reversible-deactivation radical polymerization made it possible to run controlled polymer chemistry in open labware. That is what puts it within reach of a liquid handler, and it is the premise behind our automated photo-ATRP platform, which screens ligands and initiators for monomers, methyl methacrylate among them, that are slow enough to be painful to optimise by hand. Synthesis throughput is only useful if characterisation keeps up. SAXS Assistant addresses the other end of the pipeline: it automates small-angle X-ray scattering analysis, estimates maximum particle dimension with a model trained on 1,940 experimental profiles, and refuses to report results it cannot stand behind. ## Methods ### Polymer chemistry https://emanahmed.org/research.html Polymer chemistry. Photoinduced ATRP, PET-RAFT, random copolymer synthesis, vacuum filtration, solubility characterisation ### High-throughput experimentation https://emanahmed.org/research.html High-throughput experimentation. Liquid handling robotics, 96-well plate assay design, automated enzyme activity and solubility screening ### Structural characterisation https://emanahmed.org/research.html Structural characterisation. Small-angle X-ray scattering, Guinier and P(r) analysis, Kratky interpretation, synchrotron data workflows ### Computation https://emanahmed.org/research.html Computation. Python, scikit-learn, multilayer perceptron and Gaussian mixture models, data visualisation, experimental design automation ### Biophysics https://emanahmed.org/research.html Biophysics. Protein stabilization, biocatalysis in water-miscible organic solvents, coarse-grained modelling of lipid membranes ## Publications ### Mapping Biomaterial Complexity by Machine Learning Tissue Engineering Part A · 2024 · doi:10.1089/ten.tea.2024.0067 · https://emanahmed.org/publications/mapping-biomaterial-complexity-machine-learning.html Mapping Biomaterial Complexity by Machine Learning Eman Ahmed, Prajakatta Mulay, Cesar Ramirez, Gabriela Tirado-Mansilla, Eugene Cheong, Adam J. Gormley Tissue Engineering Part A 2024 Biomaterials rarely fail or succeed for one obvious reason. Performance usually comes from a combination of small structural details interacting at once. Testing those combinations one experiment at a time does not scale. This review sets out how high-throughput experimentation paired with machine learning changes the search: run many conditions in parallel, keep every data point including the failures, and train models that map structure to function across the whole space rather than at a few sampled points. Biomaterials often have subtle properties that ultimately drive their bespoke performance. Given this nuanced structure–function behavior, the standard scientific approach of one experiment at a time or design of experiment methods is largely inefficient for the discovery of complex biomaterials. More recently, high-throughput experimentation coupled with machine learning methods has matured beyond expert users allowing scientists and engineers from diverse backgrounds to access these powerful data science tools. As a result, we now have the opportunity to strategically utilize all available data from high-throughput experiments to train efficacious models and map the structure-function behavior of biomaterials for their discovery. Herein, we discuss this necessary shift to data-driven determination of structure–function properties of biomaterials as we highlight how machine learning is leveraged in identifying physicochemical cues for biomaterials in tissue engineering, gene delivery, drug delivery, protein stabilization, and antifouling materials. We also discuss data-mining approaches that are coupled with machine learning to map biomaterial functions that reduce the load on experimental approaches for faster biomaterial discovery. Ultimately, harnessing the prowess of machine learning will lead to accelerated discovery and development of optimal biomaterial designs. ### Automation-Assisted Photoinduced Atom Transfer Radical Polymerization ACS Polymers Au · 2026 · doi:10.1021/acspolymersau.5c00067 · https://emanahmed.org/publications/automation-assisted-photo-atrp.html Automation-Assisted Photoinduced Atom Transfer Radical Polymerization Cesar Ramirez, Eman Ahmed, Elena Di Mare, Maria Pineiro-Goncalves, Apostolos Maroulis, Prajakatta Mulay, D. Christopher Radford, Adam J. Gormley ACS Polymers Au 2026 ATRP is one of the workhorse reactions for making polymers with controlled length and composition, but it has historically needed inert, oxygen-free conditions, which rules out running it in an open well plate on a robot. Oxygen-tolerant chemistry removes that constraint. This paper puts photo-ATRP onto an automated liquid-handling platform and uses it to screen reaction components at a scale that is impractical by hand, including for methyl methacrylate, a monomer that propagates slowly enough to be genuinely awkward to optimise. Oxygen-tolerant reversible-deactivation radical polymerizations (RDRP) now allow many of these reactions to proceed in open labware, such as well plates. This enables the high-throughput synthesis of tailored polymers and lowers the knowledge barrier required to obtain these materials. Building on our previous work automating photoinduced electron/energy transfer reversible addition–fragmentation chain transfer (PET–RAFT) and enzyme-assisted RAFT (Enz-RAFT) polymerization, we now introduce automated atom transfer radical polymerization (ATRP). Here, we demonstrate the potential of this platform for the high-throughput optimization of ATRP chemistry. Furthermore, we demonstrate that this workflow can help provide insights into the selection of reaction components, such as ligands and initiators, for the polymerization of kinetically difficult monomers such as methyl methacrylate with smaller rates of propagation than acrylates. This coupling paves the way for data-driven optimization of ATRP reactions, accelerated by the generation of high-throughput data sets. To facilitate the integration of robotics for high-throughput applications in polymer synthesis and optimization of photo-ATRP, we have made a Python package available to assist with experimental planning. ### SAXS Assistant: Automated SAXS analysis for structural discovery in biologics and polymeric nanoparticles Biophysical Journal · 2025 · doi:10.1016/j.bpj.2025.09.034 · https://emanahmed.org/publications/saxs-assistant-automated-saxs-analysis.html SAXS Assistant: Automated SAXS analysis for structural discovery in biologics and polymeric nanoparticles Cesar Ramirez, Elena Di Mare, James Byrnes, Eman Ahmed, Maria Pineiro-Goncalves, Cristian Lopez, N. Sanjeeva Murthy, Adam J. Gormley Biophysical Journal 2025 SAXS tells you the size and shape of something in solution, but getting there involves judgement calls: where to set the Guinier range, whether a P(r) fit is trustworthy, what maximum dimension to believe. Those calls are slow and they vary between analysts, which is a problem once you are producing hundreds of profiles. SAXS Assistant automates the pipeline, trains a model on 1,940 experimental profiles from the SASBDB to estimate Dmax, clusters profiles against known biomolecular shapes, and flags results it is not confident about instead of quietly returning them. Small-angle x-ray scattering (SAXS) is a powerful technique for assessing macromolecular structure. High-throughput SAXS is limited by the time-consuming and, at times, subjective nature of SAXS data interpretation. Here, we present SAXS Assistant, a Python-based script that streamlines SAXS data analysis to extract features for machine learning (ML) and key structural parameters, including the Guinier radius of gyration (Rg), pair distance distribution function (PDDF)-derived Rg, maximum particle dimension (Dmax), and Kratky plots. The script builds upon BioXTAS RAW and validates reliability via Guinier/PDDF Rg agreement, an important indicator of well-measured data sets. For assistance in Dmax estimation, a multilayer perceptron regressor was trained with 1940 data files from the Small Angle Scattering Biological Data Bank. The model achieved a test set performance R2 = 0.90 and mean absolute error = 11.7 Å. Training exclusively with experimental data translates analyses from researchers, including experts in the field, to the ML model, which helps assess Dmax estimations from PDDF. Gaussian mixture model clustering was implemented to classify profiles into structural classes based on entries in the Small Angle Scattering Biological Data Bank. Users may therefore assess the similarity between experimental samples and known biomolecular shapes within the mapped repository entries. This probabilistic clustering aids in quantifying information from Kratky and generating shape-descriptive features. SAXS Assistant accelerates SAXS data analysis through enforced quality control, ML-ready outputs, and flags for low-confidence results. In addition to providing the ability to analyze large data sets at high throughput, this tool is versatile and may serve researchers in both biological and synthetic polymer research fields. ### High-Throughput Approach for Evaluating the Solubility of Polymer-Stabilized Proteins in Organic Solvent https://emanahmed.org/publications.html High-Throughput Approach for Evaluating the Solubility of Polymer-Stabilized Proteins in Organic Solvent Eman Ahmed, Adam J. Gormley The doctoral project: a plate-based, automated assay for asking which random copolymers keep an enzyme soluble and active once it is moved into a water-miscible organic solvent. ## Research notes ### What oxygen tolerance buys you in polymer chemistry 2026 · https://emanahmed.org/notes/why-a-well-plate-changed-polymer-chemistry.html Controlled radical polymerisation used to require sealed, degassed glassware. Oxygen-tolerant chemistry moved it into open well plates, and that one change is what puts polymer synthesis within reach of a liquid-handling robot. If you have only ever seen controlled radical polymerisation described in a textbook, the constraint that shapes everything in practice is easy to miss: oxygen kills it. Propagating radicals react with molecular oxygen far faster than they react with monomer, so a conventional atom transfer radical polymerisation needs sealed glassware and a freeze-pump-thaw cycle or a nitrogen sparge before anything useful happens. That requirement quietly sets the ceiling on how many experiments you can run. Degassing is manual, it is slow, and it does not parallelise. A skilled person might set up a dozen carefully controlled reactions in a day. Twelve points is not enough to map a reaction space with four or five interacting variables. Open labware is the whole point Oxygen-tolerant reversible-deactivation radical polymerisation changes the arithmetic. By building oxygen consumption into the system itself, photocatalytically, enzymatically, or through the reducing environment of the photo-ATRP mechanism, the reaction becomes something you can run in an uncapped well on an open bench. Once that is true, a 96-well plate stops being a container and becomes an experimental design. Each well is an independent condition. A liquid handler can lay down ninety-six different ligand, initiator, monomer and catalyst combinations in the time it would take to degas one flask, and every one of them is measured the same way, which matters more than it sounds like it should. Consistency in how data is collected is what makes it usable for modelling later. Where it gets interesting: the slow monomers Acrylates propagate quickly and are relatively forgiving. Methacrylates are not. Methyl methacrylate has a substantially smaller propagation rate constant, which means the balance between activation and deactivation that ATRP depends on has to be tuned more carefully, and the right ligand and initiator pairing is not something you can reliably reason your way to from first principles. This is precisely the situation where throughput earns its keep. Rather than arguing about which ligand should work, you screen the ligands. In our ACS Polymers Au paper we used an automated photo-ATRP platform to do exactly that, extending a workflow the lab had previously built for PET-RAFT and enzyme-assisted RAFT to a third major polymerisation chemistry. The unglamorous part that makes it reproducible A high-throughput platform that only works in the lab that built it is a demo, not a method. The practical obstacle is rarely the robot. It is translating an intended experimental design into the deck layout, volumes and transfer sequence the instrument needs, without arithmetic errors propagating silently across ninety-six wells. We released a Python package to handle that planning step. It is not the scientifically exciting part of the work, and it is probably the part that determines whether anyone else can run it. What this unlocks The reason to care about throughput here is not throughput. It is that data-driven optimisation of polymerisation conditions needs datasets that are large, consistent and include the conditions that did not work. Hand-run chemistry produces small datasets biased towards conditions the chemist already expected to succeed. Plate-based chemistry produces the other kind. ### Trusting a SAXS analysis you didn't do by hand 2026 · https://emanahmed.org/notes/trusting-a-saxs-analysis-you-did-not-do-by-hand.html Small-angle X-ray scattering analysis is full of judgement calls that vary between analysts. Automating it is less about speed than about making those calls explicit, and about flagging the profiles where the answer should not be trusted. Small-angle X-ray scattering gives you the size and shape of something in solution without crystallising it, which is why it is so useful for proteins, polymer nanoparticles and anything else that will not sit still. What it does not give you is an unambiguous answer. Between the raw scattering curve and a reported radius of gyration sit a series of decisions, and different analysts make them differently. The judgement calls Where do you set the Guinier region? Too narrow and the fit is noisy; too wide and you violate the approximation the fit depends on. How do you choose the maximum particle dimension, Dmax, when computing the pair distance distribution function? Pick it too small and you truncate real structure, too large and you invent oscillation that is not there. Is the sample aggregating? Is there interparticle interference at low q? An experienced person handles all of this more or less automatically, and does it well. The problem is that they handle it at a rate of a few profiles an hour, and two experienced people will not produce identical numbers. Once you are generating hundreds of profiles from a high-throughput synthesis campaign, both of those facts become blocking. Cross-validation instead of assertion The design decision I think matters most in SAXS Assistant is that it does not simply report a number. There are two independent routes to a radius of gyration: the Guinier approximation at low q, and integration of the pair distance distribution function. On a well-measured dataset those two should agree. When they diverge, something is wrong: with the measurement, the buffer subtraction, or the assumption of monodispersity. Using that agreement as a built-in quality gate means the tool has a basis for declining to answer, which is the property you actually want from automation. A pipeline that confidently returns a plausible-looking number for a bad profile is worse than no pipeline. Training on what people actually did For Dmax estimation we trained a multilayer perceptron regressor on 1,940 experimental data files from the Small Angle Scattering Biological Data Bank. The choice of training data is deliberate: these are real profiles with Dmax values chosen by researchers, many of them specialists. The model is therefore learning how the field makes this judgement, not how an idealised simulation behaves. On a held-out test set it reached R² = 0.90 with a mean absolute error of 11.7 Å. That is not a replacement for expert review on a structure you intend to publish. It is entirely adequate for triaging several hundred profiles down to the handful that deserve it. Shape, without pretending to a structure The other piece is classification. We fitted a Gaussian mixture model over SASBDB entries to group scattering profiles into structural classes, so a user can ask how closely an experimental sample resembles known biomolecular shapes. The Kratky plot has always carried this information (folded, unfolded, multidomain), but reading it has been qualitative. Clustering makes it a probability rather than an impression, without overclaiming that you have solved a structure. The general point Automating an analysis forces you to write down the decisions that were previously made implicitly. That is uncomfortable, because some of them turn out not to have a principled basis. It is also the only way to make the analysis reproducible, and the flags for low-confidence results are, in the end, more valuable than the speed. ### Biomaterials discovery needs the experiments that failed 2025 · https://emanahmed.org/notes/biomaterials-discovery-needs-the-failures.html Machine learning on biomaterials data keeps running into the same problem: the published record is a survivorship-biased sample of what worked. High-throughput experimentation is one of the few ways to fix that at the source. Biomaterials are hard to design because their performance usually comes from several subtle properties interacting at once. Surface chemistry, molecular weight distribution, charge density, hydrophobic balance and architecture all contribute, and rarely independently. The structure–function relationship is real, but it is not something you can usually write down. The traditional response is to hold everything constant and vary one thing. It is rigorous, and in a space with this much interaction between variables it is close to hopeless. You sample a line through a space that has structure in every direction. The obvious fix has a less obvious problem Machine learning is the natural tool for a high-dimensional structure–function map, and the barrier to using it has dropped sharply: an experimentalist can now train a competent model without a background in statistics. That is a genuine shift, and it is the shift our review in Tissue Engineering Part A was written around. But models are only as good as what they learn from, and the biomaterials literature is a filtered sample. Papers report the formulations that worked. The ones that aggregated, or failed to release, or provoked an immune response, are mostly absent, not through dishonesty, but because a null result is hard to publish and the material was abandoned. A model trained on that record learns which successful materials resemble other successful materials. It has very little to say about where the boundary of the useful region lies, because it has never been shown the other side of it. Why high-throughput data is different in kind This is the part I think is underappreciated. When you run a 96-well plate, you keep every well. The formulations that precipitated are recorded with the same rigour as the ones that performed, because they were measured by the same instrument in the same run. You are not making a publication decision about each data point. The resulting dataset is balanced in a way that a literature-derived one structurally cannot be. That is a large part of why high-throughput experimentation and machine learning belong together, not because the throughput is impressive, but because of what it does to the shape of the data. Where data mining still helps None of which means the published record is useless. Text and data mining across existing literature can map regions of a design space cheaply enough to tell you where not to spend benchtime, and for well-studied material classes the aggregate signal is strong. Our review covers these approaches alongside direct experimentation, across tissue engineering, gene delivery, drug delivery, protein stabilization and antifouling materials. The practical position is that mined data is good for narrowing and high-throughput data is good for deciding. Used the other way round, you get a model that is confident about a region nobody has actually measured. What follows for how we run experiments If you accept that the failures carry information, some things follow. Record them with the same metadata as the successes. Do not discard a plate because most of it did not work. Report the full screened range, not the subset that supports the conclusion. And design assays so that failure produces a number rather than an absence: a solubility of zero is data; a well you stopped measuring is not.