Tommaso Biancalani

• PhD, Physics
• Group Leader at Broad Institute of MIT and Harvard

The observed single-handedness of biological amino acids and sugars has long been attributed to autocatalysis. However, the stability of homochiral states in deterministic autocatalytic systems relies on cross inhibition of the two chiral states, an unlikely scenario for early life self-replicators. Here, we present a theory for a stochastic indivi...

We present an analytical treatment of a genetic switch model consisting of two mutually inhibiting genes operating without cooperative binding of the corresponding transcription factors. Previous studies have numerically shown that these systems can exhibit bimodal dynamics without possessing two stable fixed points in the deterministic rate equati...

We investigate a type of bistability where noise not only causes transitions between stable states, but also constructs the states themselves. We focus on the experimentally well-studied system of ants choosing between two food sources to illustrate the essential points, but the ideas are more general. The mean time for switching between the two bi...

The design of regulatory elements is pivotal in gene and cell therapy, where DNA sequences are engineered to drive elevated and cell-type specific expression. However, the systematic assessment of synthetic DNA sequences without robust metrics and easy-to-use software remains challenging. Here, we introduce Polygraph, a Python framework that evalua...

Diffusion models have shown promising generative capabilities across diverse domains, yet aligning their outputs with desired reward functions remains a challenge, particularly in cases where reward functions are non-differentiable. Some gradient-free guidance methods have been developed, but they often struggle to achieve optimal inference-time al...

High-content perturbation experiments allow scientists to probe biomolecular systems at unprecedented resolution, but experimental and analysis costs pose significant barriers to widespread adoption. Machine learning has the potential to guide efficient exploration of the perturbation space and extract novel insights from these data. However, curre...

To fully leverage the capabilities of diffusion models, we are often interested in optimizing downstream reward functions during inference. While numerous algorithms for reward-guided generation have been recently proposed due to their significance, current approaches predominantly focus on single-shot generation, transitioning from fully noised to...

This tutorial provides an in-depth guide on inference-time guidance and alignment methods for optimizing downstream reward functions in diffusion models. While diffusion models are renowned for their generative modeling capabilities, practical applications in fields such as biology often require sample generation that maximizes specific metrics (e....

Predicting transcriptional responses to novel drugs provides a unique opportunity to accelerate biomedical research and advance drug discovery efforts. However, the inherent complexity and high dimensionality of cellular responses, combined with the extremely limited available experimental data, makes the task challenging. In this study, we leverag...

Single-cell RNA sequencing has profiled hundreds of millions of human cells across organs, diseases, development and perturbations to date. Mining these growing atlases could reveal cell–disease associations, identify cell states in unexpected tissue contexts and relate in vivo biology to in vitro models. These require a common measure of cell simi...

Bridging biomolecular modeling with natural language information, particularly through large language models (LLMs), has recently emerged as a promising interdisciplinary research area. LLMs, having been trained on large corpora of scientific documents, demonstrate significant potential in understanding and reasoning about biomolecules by providing...

Recent studies have demonstrated the strong empirical performance of diffusion models on discrete sequences across domains from natural language to biological sequence generation. For example, in the protein inverse folding task, conditional diffusion models have achieved impressive results in generating natural-like sequences that fold back into t...

Sequence-to-function models that predict gene expression from genomic DNA sequence have proven valuable for many biological tasks, including understanding cis-regulatory syntax and interpreting non-coding genetic variants. However, current state-of-the-art models have been trained largely on bulk expression profiles from healthy tissues or cell lin...

Deep neural networks excel in mapping genomic DNA sequences to associated readouts (e.g., protein-DNA binding). Beyond prediction, the goal of these networks is to reveal to scientists the underlying motifs (and their syntax) which drive genome regulation. Traditional methods that extract motifs from convolutional filters suffer from the uninterpre...

The continued scaling of genetic perturbation technologies combined with high-dimensional assays such as cellular microscopy and RNA-sequencing has enabled genome-scale reverse-genetics experiments that go beyond single-endpoint measurements of growth or lethality. Datasets emerging from these experiments can be combined to construct perturbative “...

Cis -regulatory elements (CREs), such as promoters and enhancers, are DNA sequences that regulate the expression of genes. The activity of a CRE is influenced by the order, composition, and spacing of sequence motifs that are bound by proteins called transcription factors (TFs). Synthetic CREs with specific properties are needed for biomanufacturin...

Deep learning models are increasingly being used to perform a variety of tasks on DNA sequences, such as predicting tissue- and cell type- specific sequence activity, deriving cis-regulatory rules, predicting non-coding variant effects, and designing synthetic regulatory sequences. However, these models require specialized knowledge to build, train...

The proliferation of multi-drug-resistant bacteria underscores an urgent need for novel antibiotics. Traditional discovery methods face challenges due to limited chemical diversity, high costs, and difficulties in identifying structurally novel compounds. Here, we explore the integration of small molecule high-throughput screening with a deep learn...

Dendritic cell (DC) homeostasis is maintained in secondary lymphoid organs (SLOs) by Fms-like tyrosine kinase 3 ligand (FLT3L). The specific niche providing this DC growth factor within human and mouse SLOs is unclear. Here, we show that Gremlin1 (Grem1)-expressing lymph node fibroblastic reticular cells (FRCs) support DC homeostasis via provision...

Diffusion models excel at capturing the natural design spaces of images, molecules, DNA, RNA, and protein sequences. However, rather than merely generating designs that are natural, we often aim to optimize downstream reward functions while preserving the naturalness of these design spaces. Existing methods for achieving this goal often require ``d...

Computational and machine learning approaches to model the conformational landscape of macrocyclic peptides have the potential to enable rational design and optimization. However, accurate, fast, and scalable methods for modeling macrocycle geometries remain elusive. Recent deep learning approaches have significantly accelerated protein structure p...

This tutorial provides a comprehensive survey of methods for fine-tuning diffusion models to optimize downstream reward functions. While diffusion models are widely known to provide excellent generative modeling capability, practical applications in domains such as biology require generating samples that maximize some desired metric (e.g., translat...

Diffusion models are powerful generative models that allow for precise control over the characteristics of the generated samples. While these diffusion models trained on large datasets have achieved success, there is often a need to introduce additional controls in downstream fine-tuning processes, treating these powerful models as pre-trained diff...

AI-driven design problems, such as DNA/protein sequence design, are commonly tackled from two angles: generative modeling, which efficiently captures the feasible design space (e.g., natural images or biological sequences), and model-based optimization, which utilizes reward models for extrapolation. To combine the strengths of both approaches, we...

Single-cell RNA sequencing and other profiling assays have helped interrogate cells at unprecedented resolution and scale, but are inherently destructive. Raman microscopy reports on the vibrational energy levels of proteins and metabolites in a label-free and nondestructive manner at subcellular spatial resolution, but it lacks genetic and molecul...

Effective tissue repair requires coordinated intercellular communication to sense damage, remodel the tissue, and restore function. Here, we dissected the healing response in the intestinal mucosa by mapping intercellular communication at single-cell resolution and integrating with spatial transcriptomics. We demonstrated that a risk variant for Cr...

Single-cell RNA-seq (scRNA-seq) studies have profiled over 100 million human cells across diseases, developmental stages, and perturbations to date. A singular view of this vast and growing expression landscape could help reveal novel associations between cell states and diseases, discover cell states in unexpected tissue contexts, and relate in vi...

With the advent of multiplex fluorescence in situ hybridization (FISH) and in situ RNA sequencing technologies, spatial transcriptomics analysis is advancing rapidly, providing spatial location and gene expression information about cells in tissue sections at single cell resolution. Cell type classification of these spatially-resolved cells can be...

Diffusion models have achieved state-of-the-art performance in generating many different kinds of data, including images, text, and videos. Despite their success, there has been limited research on how the underlying diffusion process and the final convergent prior can affect generative performance; this research has also been limited to continuous...

Macrocyclic peptides are an emerging therapeutic modality, yet computational approaches for accurately sampling their diverse 3D ensembles remain challenging due to their conformational diversity and geometric constraints. Here, we introduce RINGER, a diffusion-based transformer model for sequence-conditioned generation of macrocycle structures bas...

Computational and machine learning approaches to model the conformational landscape of macrocyclic peptides have the potential to enable rational design and optimization. However, accurate, fast, and scalable methods for modeling macrocycle geometries remain elusive. Recent deep learning approaches have significantly accelerated protein structure p...

Generative flow networks (GFlowNets) are a family of algorithms that learn a generative policy to sample discrete objects $x$ with non-negative reward $R(x)$. Learning objectives guarantee the GFlowNet samples $x$ from the target distribution $p^*(x) \propto R(x)$ when loss is globally minimized over all states or trajectories, but it is unclear ho...

Learning causal relationships between variables is a well-studied problem in statistics, with many important applications in science. However, modeling real-world systems remain challenging , as most existing algorithms assume that the underlying causal graph is acyclic. While this is a convenient framework for developing theoretical developments a...

Diffusion models achieve state-of-the-art performance in generating realistic objects and have been successfully applied to images, text, and videos. Recent work has shown that diffusion can also be defined on graphs, including graph representations of drug-like molecules. Unfortunately, it remains difficult to perform conditional generation on gra...

Deep graph generative modeling has proven capable of learning the distribution of complex, multi-scale structures characterizing real-world graphs. However, one of the main limitations of existing methods is their large output space, which limits generation scalability and hinders accurate modeling of the underlying distribution. To overcome these...

Diffusion models have achieved justifiable popularity by attaining state-of-the-art performance in generating realistic objects from seemingly arbitrarily complex data distributions, including when conditioning generation on labels. Unfortunately, however, their iterative nature renders them very computationally inefficient during the sampling proc...

Molecular shape and geometry dictate key biophysical recognition processes, yet many graph neural networks disregard 3D information for molecular property prediction. Here, we propose a new contrastive-learning procedure for graph neural networks, Molecular Contrastive Learning from Shape Similarity (MolCLaSS), that implicitly learns a three-dimens...

How much explicit guidance is necessary for conditional diffusion? We consider the problem of conditional sampling using an unconditional diffusion model and limited explicit guidance (e.g., a noised classifier, or a conditional diffusion model) that is restricted to a small number of time steps. We explore a model predictive control (MPC)-like app...

With the advent of multiplex fluorescence in situ hybridization (FISH) and in situ RNA sequencing technologies, spatial transcriptomics analysis is advancing rapidly. Spatial transcriptomics provides spatial location and pattern information about cells in tissue sections at single cell resolution. Cell type classification of spatially-resolved cell...

jats:p>Establishing causal relationships between genetic alterations of human cancers and specific phenotypes of malignancy remains a challenge. We sequentially introduced mutations into healthy human melanocytes in up to five genes spanning six commonly disrupted melanoma pathways, forming nine genetically distinct cellular models of melanoma. We...

Single cell RNA-Seq (scRNA-seq) and other profiling assays have opened new windows into understanding the properties, regulation, dynamics, and function of cells at unprecedented resolution and scale. However, these assays are inherently destructive, precluding us from tracking the temporal dynamics of live cells, in cell culture or whole organisms...

Charting an organs’ biological atlas requires us to spatially resolve the entire single-cell transcriptome, and to relate such cellular features to the anatomical scale. Single-cell and single-nucleus RNA-seq (sc/snRNA-seq) can profile cells comprehensively, but lose spatial information. Spatial transcriptomics allows for spatial measurements, but...

Single-cell transcriptomics can provide quantitative molecular signatures for large, unbiased samples of the diverse cell types in the brain1–3. With the proliferation of multi-omics datasets, a major challenge is to validate and integrate results into a biological understanding of cell-type organization. Here we generated transcriptomes and epigen...

Here we report the generation of a multimodal cell census and atlas of the mammalian primary motor cortex as the initial product of the BRAIN Initiative Cell Census Network (BICCN). This was achieved by coordinated large-scale analyses of single-cell transcriptomes, chromatin accessibility, DNA methylomes, spatially resolved single-cell transcripto...

Single-cell RNA sequencing has revealed extensive transcriptional cell state diversity in cancer, often observed independently of genetic heterogeneity, raising the central question of how malignant cell states are encoded epigenetically. To address this, here we performed multiomics single-cell profiling—integrating DNA methylation, transcriptome...

The mammalian cerebral cortex has an unparalleled diversity of cell types, which are generated during development through a series of temporally orchestrated events that are under tight evolutionary constraint and are critical for proper cortical assembly and function1,2. However, the molecular logic that governs the establishment and organization...

Patients with chronic lung disease (CLD) have an increased risk for severe coronavirus disease-19 (COVID-19) and poor outcomes. Here, we analyze the transcriptomes of 611,398 single cells isolated from healthy and CLD lungs to identify molecular characteristics of lung cells that may account for worse COVID-19 outcomes in patients with chronic lung...

Human diffuse gliomas are incurable malignancies, where cellular state diversity fuels tumor progression and resistance to therapy. Single-cell RNA-sequencing (scRNAseq) studies recently charted the cellular states of the two major categories of human gliomas, IDH-mutant gliomas (IDH-MUT) and IDH-wildtype glioblastoma (GBM), showing that malignant...

Human diffuse gliomas are incurable brain tumors, where cellular state diversity fuels tumor progression and resistance to therapy. Single-cell RNA-sequencing (scRNAseq) studies recently charted the cellular states of the two major categories of human gliomas, IDH-mutant gliomas (IDH-MUT) and IDH-wildtype glioblastoma (GBM), showing that malignant...

We report the generation of a multimodal cell census and atlas of the mammalian primary motor cortex (MOp or M1) as the initial product of the BRAIN Initiative Cell Census Network (BICCN). This was achieved by coordinated large-scale analyses of single-cell transcriptomes, chromatin accessibility, DNA methylomes, spatially resolved single-cell tran...

Charting a biological atlas of an organ, such as the brain, requires us to spatially-resolve whole transcriptomes of single cells, and to relate such cellular features to the histological and anatomical scales. Single-cell and single-nucleus RNA-Seq (sc/snRNA-seq) can map cells comprehensively, but relating those to their histological and anatomica...

Single cell transcriptomics has transformed the characterization of brain cell identity by providing quantitative molecular signatures for large, unbiased samples of brain cell populations. With the proliferation of taxonomies based on individual datasets, a major challenge is to integrate and validate results toward defining biologically meaningfu...

Single cell transcriptomics has transformed the characterization of brain cell identity by providing quantitative molecular signatures for large, unbiased samples of brain cell populations. With the proliferation of taxonomies based on individual datasets, a major challenge is to integrate and validate results toward defining biologically meaningfu...

Understanding the genetic and molecular drivers of phenotypic heterogeneity across individuals is central to biology. As new technologies enable fine-grained and spatially resolved molecular profiling, we need new computational approaches to integrate data from the same organ across different individuals into a consistent reference and to construct...

Transformative technologies are enabling the construction of three-dimensional maps of tissues with unprecedented spatial and molecular resolution. Over the next seven years, the NIH Common Fund Human Biomolecular Atlas Program (HuBMAP) intends to develop a widely accessible framework for comprehensively mapping the human body at single-cell resolu...

Abstract Quantifying virulence remains a central problem in human health, pest control, disease ecology, and evolutionary biology. Bacterial virulence is typically quantified by the LT50 (i.e., the time taken to kill 50% of infected hosts); however, such an indicator cannot account for the full complexity of the infection process, such as distingui...

Quantifying virulence remains a central problem in human health, pest control, disease ecology, and evolutionary biology. Bacterial virulence is typically quantified by the LT50 (i.e. the time taken to kill 50% of infected hosts), however, such an indicator cannot account for the full complexity of the infection process, such as distinguishing betw...

All known life on the Earth exhibits at least two non-trivial common features: the canonical genetic code and biological homochirality, both of which emerged prior to the Last Universal Common Ancestor state. This article describes recent efforts to provide a narrative of this epoch using tools from statistical mechanics. During the emergence of se...

We consider a stochastic version of the Wilson–Cowan model which accommodates for discrete populations of excitatory and inhibitory neurons. The model assumes a finite carrying capacity with the two populations being constant in size. The master equation that governs the dynamics of the stochastic model is analyzed by an expansion in powers of the...

[This corrects the article DOI: 10.1371/journal.pbio.1002540.].

Excel file containing the data used to create Figs 1–5 and S1–S5 Figs. (XLSX)

The origin of homochirality, the observed single-handedness of biological amino acids and sugars, has long been attributed to autocatalysis, a frequently assumed precursor for early life self-replication. However, the stability of homochiral states in deterministic autocatalytic systems relies on cross-inhibition of the two chiral states, an unlike...

Virtually every interesting natural phenomenon, not least life itself, entails physical systems being forced to flow thermodynamically up-hill, away from equilibrium rather than towards it. This requires the action of a mechanism, acting as an ëngine”, which lashes the up-hill process to a more powerful one proceeding in its spontaneous, down-hill...

We consider a stochastic version of the Wilson-Cowan model which accommodates for discrete populations of excitatory and inhibitory neurons. The model assumes a finite carrying capacity with the two populations being constant in size. The master equation that governs the dynamics of the stochastic model is analyzed by an expansion in powers of the...

Mutualisms between species play an important role in ecosystem function and stability. However, in some environments, the competitive aspects of an interaction may dominate the mutualistic aspects. Although these transitions could have far-reaching implications, it has been difficult to study the causes and consequences of this mutualistic–competit...

Relative fitness of Leu- is lower than fitness of Trp-. To analyze relative fitness, we grew the strains in co-culture at saturating amino acid concentrations (200 μM tryptophan and 1600 μM leucine). With such high concentrations, additional amino acids provided through cross-feeding will give negligible benefits, thus enabling us to compare the in...

Relative fitness as a function of relative abundance. To determine equilibria in co-cultures that had not yet reached saturation, we determined relative fitness as a function of the fraction of Leu- cells. Co-cultures were grown for 2 d to reach carrying capacity, after which relative fitness was determined as described earlier (S1 Information). Re...

Smaller niche overlap results in a larger region of mutualistic interactions. Simulations were run to determine qualitative interaction as a function of supplemented amino acids (a) and niche overlap (c). Niche overlap was modelled as the degree to which each strain affects the carrying capacity of the other strain (S1 Information), with c = 1 bein...

Order of qualitative interactions is robust for changing death rates. Simulations were run to determine qualitative interactions as a function of supplemented amino acids (a) and death rate (δ). The model shifts through the same order of qualitative interactions in a large range of death rates. (PDF)

Excel file containing the data used to create Figs 1–5 and S1–S5 Figs. (XLSX)

Individual tracks of co-culture abundances at different amino acid concentrations. Plots show individuals traces of experiments used for Fig 3. Co-cultures were grown at 16 different amino acid concentrations, ranging from 0 μM tryptophan and 0 μM leucine to 200 μM tryptophan and 1,600 μM leucine. Co-cultures were started at six different relative...

Growth curves of S. cerevisiae strains before and after adaptation to low amino acids. Trp- (A) and Leu- (B) cells in exponential phase were seeded in 96-well flat bottom plates and incubated at 30°C for 32 h. Density was measured automatically every 10 min through spectrophotometry. Cells were either adapted (dashed lines) or not adapted (solid li...

Zip file containing the following files: “SM_analytical_treatment.nb” is a Wolfram Mathematica notebook file that contains the code used to generate Figs 2 and 4. “SM_analytical_treatment.cdf” contains the same code, but for the freely available software CDF interactive player. “SM_analytical_treatment.pdf” contains the PDF version and the figures...

Different ecological regimes are revealed by bifurcation analysis of the model. Insets I–V are phase portraits of Eqs 1 and 2 obtained for various values of a (other parameters values given in the main text). Eigenvectors have normalized length. Inset I: (a = 0.08, extinction) all trajectories are attracted to global extinction. Inset II: (a = 0.09...

Supplementary information on data analysis and modelling. (PDF)

A cross-feeding mutualism can protect against invasion by other strains. Plot shows equilibrium density of simulations with four strains as a function of supplemented amino acids. Double producers (yellow line) are modelled to have a lower growth rate than single producers (red and green lines, equivalent to strain X and Y in Eqs 1 and 2), whereas...

The amplitude of fluctuation-induced patterns might be expected to be proportional to the strength of the driving noise, suggesting that such patterns would be difficult to observe in nature. Here, we show that a large class of spatially-extended dynamical systems driven by intrinsic noise can exhibit giant amplification, yielding patterns whose am...

The amplitude of fluctuation-induced patterns might be expected to be proportional to the strength of the driving noise, suggesting that such patterns would be difficult to observe in nature. Here, we show that a large class of spatially-extended dynamical systems driven by intrinsic noise can exhibit giant amplification, yielding patterns whose am...

DOI:https://doi.org/10.1103/PhysRevE.92.039902

We develop an theoretical approach for predicting biodiversity in multi-dimensional niche spaces, arising due to ecological drivers such as competitive exclusion. The novelty of our approach relies on the fact that ecological niches are described by sequences of strings, which allows us to describe multiple traits. We define the mathematical framew...

We develop an theoretical approach for predicting biodiversity in multi-dimensional niche spaces, arising due to ecological drivers such as competitive exclusion. The novelty of our approach relies on the fact that ecological niches are described by sequences of strings, which allows us to describe multiple traits. We define the mathematical framew...

We investigate the statistics of the time taken for a system driven by recruitment to reach fixation. Our model describes a series of experiments where a population is confronted with two identical options, resulting in the system fixating on one of the options. For a specific population size, we show that the time distribution behaves like an inve...

We here review in more detail the mathematical formalism underlying the modelling of stochasticity in population systems, which will be used throughout the rest of the thesis. We begin with revisiting the basic mathematical definitions which lead to the concept of homogeneous stochastic process. Those describe the dynamics of the chemical concentra...

In order to study the phenomenon of noise-induced bistability we have expanded the master equation with respect to the cell volume, and obtained an approximated stochastic differential equation which was more amenable to analytical manipulations. As already remarked, the technique we have used made it possible to capture the multiplicative nature o...

The investigation of the stochastic waves in the previous chapter, has involved an analysis of a spatially extended model in which space was represented as a regular lattice.

Bistability indicates the ability of a system to reside in either of two states. This is a fairly general paradigm that has found numerous applications in science, from liquid crystal displays to the lac operon in E. coli to the cell cycle oscillator in Xenopus laevis.

Stochastic reaction-diffusion models can be analytically studied on complex networks using the linear noise approximation. This is illustrated through the use of a specific stochastic model, which displays traveling waves in its deterministic limit. The role of stochastic fluctuations is investigated and shown to drive the emergence of stochastic w...

Diffusion of a two component fluid is studied in the framework of differential equations, but where these equations are systematically derived from a well-defined microscopic model. The model has a finite carrying capacity imposed upon it at the mesoscopic level and this is shown to lead to nonlinear cross diffusion terms that modify the convention...