Yi Qin Gao

Sirui Liu, Jun Zhang, Haotian Chu et al.

Proteins are essential component of human life and their structures are important for function and mechanism analysis. Recent work has shown the potential of AI-driven methods for protein structure prediction. However, the development of new models is restricted by the lack of dataset and benchmark training procedure. To the best of our knowledge, the existing open source datasets are far less to satisfy the needs of modern protein sequence-structure related research. To solve this problem, we present the first million-level protein structure prediction dataset with high coverage and diversity, named as PSP. This dataset consists of 570k true structure sequences (10TB) and 745k complementary distillation sequences (15TB). We provide in addition the benchmark training procedure for SOTA protein structure prediction model on this dataset. We validate the utility of this dataset for training by participating CAMEO contest in which our model won the first place. We hope our PSP dataset together with the training benchmark can enable a broader community of AI/biology researchers for AI-driven protein related research.

Zhengcheng Shen, Yi Gao, Linh Kästner et al.

The advancement of computer vision and machine learning has made datasets a crucial element for further research and applications. However, the creation and development of robots with advanced recognition capabilities are hindered by the lack of appropriate datasets. Existing image or video processing datasets are unable to accurately depict observations from a moving robot, and they do not contain the kinematics information necessary for robotic tasks. Synthetic data, on the other hand, are cost-effective to create and offer greater flexibility for adapting to various applications. Hence, they are widely utilized in both research and industry. In this paper, we propose the dataset HabitatDyn, which contains both synthetic RGB videos, semantic labels, and depth information, as well as kinetics information. HabitatDyn was created from the perspective of a mobile robot with a moving camera, and contains 30 scenes featuring six different types of moving objects with varying velocities. To demonstrate the usability of our dataset, two existing algorithms are used for evaluation and an approach to estimate the distance between the object and camera is implemented based on these segmentation methods and evaluated through the dataset. With the availability of this dataset, we aspire to foster further advancements in the field of mobile robotics, leading to more capable and intelligent robots that can navigate and interact with their environments more effectively. The code is publicly available at https://github.com/ignc-research/HabitatDyn.

YuPeng Huang, Hong Zhang, Siyuan Jiang et al.

Virtual screening, including molecular docking, plays an essential role in drug discovery. Many traditional and machine-learning based methods are available to fulfil the docking task. The traditional docking methods are normally extensively time-consuming, and their performance in blind docking remains to be improved. Although the runtime of docking based on machine learning is significantly decreased, their accuracy is still limited. In this study, we take the advantage of both traditional and machine-learning based methods, and present a method Deep Site and Docking Pose (DSDP) to improve the performance of blind docking. For the traditional blind docking, the entire protein is covered by a cube, and the initial positions of ligands are randomly generated in the cube. In contract, DSDP can predict the binding site of proteins and provide an accurate searching space and initial positions for the further conformational sampling. The docking task of DSDP makes use of the score function and a similar but modified searching strategy of AutoDock Vina, accelerated by implementation in GPUs. We systematically compare its performance with the state-of-the-art methods, including Autodock Vina, GNINA, QuickVina, SMINA, and DiffDock. DSDP reaches a 29.8% top-1 success rate (RMSD < 2 Å) on an unbiased and challenging test dataset with 1.2 s wall-clock computational time per system. Its performances on DUD-E dataset and the time-split PDBBind dataset used in EquiBind, TankBind, and DiffDock are also effective, presenting a 57.2% and 41.8% top-1 success rate with 0.8 s and 1.0 s per system, respectively.

Jun Zhang, Sirui Liu, Mengyun Chen et al.

Data-driven predictive methods which can efficiently and accurately transform protein sequences into biologically active structures are highly valuable for scientific research and medical development. Determining accurate folding landscape using co-evolutionary information is fundamental to the success of modern protein structure prediction methods. As the state of the art, AlphaFold2 has dramatically raised the accuracy without performing explicit co-evolutionary analysis. Nevertheless, its performance still shows strong dependence on available sequence homologs. Based on the interrogation on the cause of such dependence, we presented EvoGen, a meta generative model, to remedy the underperformance of AlphaFold2 for poor MSA targets. By prompting the model with calibrated or virtually generated homologue sequences, EvoGen helps AlphaFold2 fold accurately in low-data regime and even achieve encouraging performance with single-sequence predictions. Being able to make accurate predictions with few-shot MSA not only generalizes AlphaFold2 better for orphan sequences, but also democratizes its use for high-throughput applications. Besides, EvoGen combined with AlphaFold2 yields a probabilistic structure generation method which could explore alternative conformations of protein sequences, and the task-aware differentiable algorithm for sequence generation will benefit other related tasks including protein design.

Yi Gao, Chenwei Tang, Jiancheng Lv

Generalized Zero-Shot Learning (GZSL) aims to recognize both seen and unseen classes by training only the seen classes, in which the instances of unseen classes tend to be biased towards the seen class. In this paper, we propose a Cluster-based Contrastive Disentangling (CCD) method to improve GZSL by alleviating the semantic gap and domain shift problems. Specifically, we first cluster the batch data to form several sets containing similar classes. Then, we disentangle the visual features into semantic-unspecific and semantic-matched variables, and further disentangle the semantic-matched variables into class-shared and class-unique variables according to the clustering results. The disentangled learning module with random swapping and semantic-visual alignment bridges the semantic gap. Moreover, we introduce contrastive learning on semantic-matched and class-unique variables to learn high intra-set and intra-class similarity, as well as inter-set and inter-class discriminability. Then, the generated visual features conform to the underlying characteristics of general images and have strong discriminative information, which alleviates the domain shift problem well. We evaluate our proposed method on four datasets and achieve state-of-the-art results in both conventional and generalized settings.

Chenwei Tang, Wenqiang Zhou, Dong Wang et al.

The advent of Industry 4.0 has precipitated the incorporation of Artificial Intelligence (AI) methods within industrial contexts, aiming to realize intelligent manufacturing, operation as well as maintenance, also known as industrial intelligence. However, intricate industrial milieus, particularly those relating to energy exploration and production, frequently encompass data characterized by long-tailed class distribution, sample imbalance, and domain shift. These attributes pose noteworthy challenges to data-centric Deep Learning (DL) techniques, crucial for the realization of industrial intelligence. The present study centers on the intricate and distinctive industrial scenarios of Nuclear Power Generation (NPG), meticulously scrutinizing the application of DL techniques under the constraints of finite data samples. Initially, the paper expounds on potential employment scenarios for AI across the full life-cycle of NPG. Subsequently, we delve into an evaluative exposition of DL's advancement, grounded in the finite sample perspective. This encompasses aspects such as small-sample learning, few-shot learning, zero-shot learning, and open-set recognition, also referring to the unique data characteristics of NPG. The paper then proceeds to present two specific case studies. The first revolves around the automatic recognition of zirconium alloy metallography, while the second pertains to open-set recognition for signal diagnosis of machinery sensors. These cases, spanning the entirety of NPG's life-cycle, are accompanied by constructive outcomes and insightful deliberations. By exploring and applying DL methodologies within the constraints of finite sample availability, this paper not only furnishes a robust technical foundation but also introduces a fresh perspective toward the secure and efficient advancement and exploitation of this advanced energy source.

Jianhua Zhang, Yi Gao, Ruyu Liu et al.

Knowledge distillation (KD) is a model compression technique that transfers knowledge from a large teacher model to a smaller student model to enhance its performance. Existing methods often assume that the student model is inherently inferior to the teacher model. However, we identify that the fundamental issue affecting student performance is the bias transferred by the teacher. Current KD frameworks transmit both right and wrong knowledge, introducing bias that misleads the student model. To address this issue, we propose a novel strategy to rectify bias and greatly improve the student model's performance. Our strategy involves three steps: First, we differentiate knowledge and design a bias elimination method to filter out biases, retaining only the right knowledge for the student model to learn. Next, we propose a bias rectification method to rectify the teacher model's wrong predictions, fundamentally addressing bias interference. The student model learns from both the right knowledge and the rectified biases, greatly improving its prediction accuracy. Additionally, we introduce a dynamic learning approach with a loss function that updates weights dynamically, allowing the student model to quickly learn right knowledge-based easy tasks initially and tackle hard tasks corresponding to biases later, greatly enhancing the student model's learning efficiency. To the best of our knowledge, this is the first strategy enabling the student model to surpass the teacher model. Experiments demonstrate that our strategy, as a plug-and-play module, is versatile across various mainstream KD frameworks. We will release our code after the paper is accepted.

Shengyang Wu, Yi Gao

As one of the most extensive social networking services, Twitter has more than 300 million active users as of 2022. Among its many functions, Twitter is now one of the go-to platforms for consumers to share their opinions about products or experiences, including flight services provided by commercial airlines. This study aims to measure customer satisfaction by analyzing sentiments of Tweets that mention airlines using a machine learning approach. Relevant Tweets are retrieved from Twitter's API and processed through tokenization and vectorization. After that, these processed vectors are passed into a pre-trained machine learning classifier to predict the sentiments. In addition to sentiment analysis, we also perform lexical analysis on the collected Tweets to model keywords' frequencies, which provide meaningful contexts to facilitate the interpretation of sentiments. We then apply time series methods such as Bollinger Bands to detect abnormalities in sentiment data. Using historical records from January to July 2022, our approach is proven to be capable of capturing sudden and significant changes in passengers' sentiment. This study has the potential to be developed into an application that can help airlines, along with several other customer-facing businesses, efficiently detect abrupt changes in customers' sentiments and take adequate measures to counteract them.

Xiangyu Wu, Yi Gao, Hailiang Zhang et al.

In this paper, we present our solution to the New frontiers for Zero-shot Image Captioning Challenge. Different from the traditional image captioning datasets, this challenge includes a larger new variety of visual concepts from many domains (such as COVID-19) as well as various image types (photographs, illustrations, graphics). For the data level, we collect external training data from Laion-5B, a large-scale CLIP-filtered image-text dataset. For the model level, we use OFA, a large-scale visual-language pre-training model based on handcrafted templates, to perform the image captioning task. In addition, we introduce contrastive learning to align image-text pairs to learn new visual concepts in the pre-training stage. Then, we propose a similarity-bucket strategy and incorporate this strategy into the template to force the model to generate higher quality and more matching captions. Finally, by retrieval-augmented strategy, we construct a content-rich template, containing the most relevant top-k captions from other image-text pairs, to guide the model in generating semantic-rich captions. Our method ranks first on the leaderboard, achieving 105.17 and 325.72 Cider-Score in the validation and test phase, respectively.

Sihao Yuan, Xu Han, Jun Zhang et al.

Theoretical studies on chemical reaction mechanisms have been crucial in organic chemistry. Traditionally, calculating the manually constructed molecular conformations of transition states for chemical reactions using quantum chemical calculations is the most commonly used method. However, this way is heavily dependent on individual experience and chemical intuition. In our previous study, we proposed a research paradigm that uses enhanced sampling in molecular dynamics simulations to study chemical reactions. This approach can directly simulate the entire process of a chemical reaction. However, the computational speed limits the use of high-precision potential energy functions for simulations. To address this issue, we present a scheme for training high-precision force fields for molecular modeling using a previously developed graph-neural-network-based molecular model, molecular configuration transformer. This potential energy function allows for highly accurate simulations at a low computational cost, leading to more precise calculations of the mechanism of chemical reactions. We applied this approach to study a Claisen rearrangement reaction and a Carbonyl insertion reaction catalyzed by Manganese.

Zicheng Ma, Chuanliu Fan, Zhicong Wang et al.

Large language models have made remarkable progress in the field of molecular science, particularly in understanding and generating functional small molecules. This success is largely attributed to the effectiveness of molecular tokenization strategies. In protein science, the amino acid sequence serves as the sole tokenizer for LLMs. However, many fundamental challenges in protein science are inherently structure-dependent. The absence of structure-aware tokens significantly limits the capabilities of LLMs for comprehensive biomolecular comprehension and multimodal generation. To address these challenges, we introduce a novel framework, ProtTeX, which tokenizes the protein sequences, structures, and textual information into a unified discrete space. This innovative approach enables joint training of the LLM exclusively through the Next-Token Prediction paradigm, facilitating multimodal protein reasoning and generation. ProtTeX enables general LLMs to perceive and process protein structures through sequential text input, leverage structural information as intermediate reasoning components, and generate or manipulate structures via sequential text output. Experiments demonstrate that our model achieves significant improvements in protein function prediction, outperforming the state-of-the-art domain expert model with a twofold increase in accuracy. Our framework enables high-quality conformational generation and customizable protein design. For the first time, we demonstrate that by adopting the standard training and inference pipelines from the LLM domain, ProtTeX empowers decoder-only LLMs to effectively address diverse spectrum of protein-related tasks.

Chuanliu Fan, Zicheng Ma, Jun Gao et al.

Recent advances in protein large language models, such as ProtTeX, represent both side-chain amino acids and backbone structure as discrete token sequences of residue length. While this design enables unified modeling of multimodal protein information, it suffers from two major limitations: (1) The concatenation of sequence and structure tokens approximately doubles the protein length and breaks the intrinsic residue-level alignment between modalities. (2) Constrained by the training corpus and limited context window, ProtTeX is typically trained on single-protein inputs, rendering it incompatible with in-context learning (ICL) and thus limiting its generalization capability. To address these issues, we propose ProtTeX-CC, a lightweight two-stage compression framework designed to enhance ProtTeX under few-shot settings. We first design a joint embedding compression mechanism that fuses sequence and structure representations at the residue level, effectively reducing the protein input length by half without sacrificing performance. Then we propose a self-compression module that aggregates each full demonstration into the latent space of the last few linguistic tokens, reducing the average demonstration length from 751 tokens to less than 16 tokens. Compared to the original ProtTeX, our self-compression approach achieves a compression ratio of approximately 93.68% in the total prompt length under the 16-shot setting. Without modifying the backbone model, ProtTeX-CC introduces only a small number of additional parameters through PEFT-based tuning in the joint embedding compression stage and a single trainable projection layer in the self-compression stage. Extensive experiments on protein function prediction show that ProtTeX-CC improves performance on the in-domain benchmark by 2%, and generalizes well to the out-of-domain dataset with a performance gain of 11%.

Mu Tian, Qinzhu Yang, Yi Gao

The success of deep networks in medical image segmentation relies heavily on massive labeled training data. However, acquiring dense annotations is a time-consuming process. Weakly-supervised methods normally employ less expensive forms of supervision, among which scribbles started to gain popularity lately thanks to its flexibility. However, due to lack of shape and boundary information, it is extremely challenging to train a deep network on scribbles that generalizes on unlabeled pixels. In this paper, we present a straightforward yet effective scribble supervised learning framework. Inspired by recent advances of transformer based segmentation, we create a pluggable spatial self-attention module which could be attached on top of any internal feature layers of arbitrary fully convolutional network (FCN) backbone. The module infuses global interaction while keeping the efficiency of convolutions. Descended from this module, we construct a similarity metric based on normalized and symmetrized attention. This attentive similarity leads to a novel regularization loss that imposes consistency between segmentation prediction and visual affinity. This attentive similarity loss optimizes the alignment of FCN encoders, attention mapping and model prediction. Ultimately, the proposed FCN+Attention architecture can be trained end-to-end guided by a combination of three learning objectives: partial segmentation loss, a customized masked conditional random fields and the proposed attentive similarity loss. Extensive experiments on public datasets (ACDC and CHAOS) showed that our framework not just out-performs existing state-of-the-art, but also delivers close performance to fully-supervised benchmark. Code will be available upon publication.

Jun Zhang, Xiaohan Lin, Weinan E et al.

Multiscale molecular modeling is widely applied in scientific research of molecular properties over large time and length scales. Two specific challenges are commonly present in multiscale modeling, provided that information between the coarse and fine representations of molecules needs to be properly exchanged: One is to construct coarse grained models by passing information from the fine to coarse levels; the other is to restore finer molecular details given coarse grained configurations. Although these two problems are commonly addressed independently, in this work, we present a theory connecting them, and develop a methodology called Cycle Coarse Graining (CCG) to solve both problems in a unified manner. In CCG, reconstruction can be achieved via a tractable deep generative model, allowing retrieval of fine details from coarse-grained simulations. The reconstruction in turn delivers better coarse-grained models which are informed of the fine-grained physics, and enables calculation of the free energies in a rare-event-free manner. CCG thus provides a systematic way for multiscale molecular modeling, where the finer details of coarse-grained simulations can be efficiently retrieved, and the coarse-grained models can be improved consistently.

Jun Zhang, Yao-Kun Lei, Yaqiang Zhou et al.

Deep learning is changing many areas in molecular physics, and it has shown great potential to deliver new solutions to challenging molecular modeling problems. Along with this trend arises the increasing demand of expressive and versatile neural network architectures which are compatible with molecular systems. A new deep neural network architecture, Molecular Configuration Transformer (Molecular CT), is introduced for this purpose. Molecular CT is composed of a relation-aware encoder module and a computationally universal geometry learning unit, thus able to account for the relational constraints between particles meanwhile scalable to different particle numbers and invariant with respect to the trans-rotational transforms. The computational efficiency and universality make Molecular CT versatile for a variety of molecular learning scenarios and especially appealing for transferable representation learning across different molecular systems. As examples, we show that Molecular CT enables representational learning for molecular systems at different scales, and achieves comparable or improved results on common benchmarks using a more light-weighted structure compared to baseline models.

Jun Zhang, Yao-Kun Lei, Zhen Zhang et al.

Deep learning is transforming many areas in science, and it has great potential in modeling molecular systems. However, unlike the mature deployment of deep learning in computer vision and natural language processing, its development in molecular modeling and simulations is still at an early stage, largely because the inductive biases of molecules are completely different from those of images or texts. Footed on these differences, we first reviewed the limitations of traditional deep learning models from the perspective of molecular physics, and wrapped up some relevant technical advancement at the interface between molecular modeling and deep learning. We do not focus merely on the ever more complex neural network models, instead, we emphasize the theories and ideas behind modern deep learning. We hope that transacting these ideas into molecular modeling will create new opportunities. For this purpose, we summarized several representative applications, ranging from supervised to unsupervised and reinforcement learning, and discussed their connections with the emerging trends in deep learning. Finally, we outlook promising directions which may help address the existing issues in the current framework of deep molecular modeling.

Jun Zhang, Yao-Kun Lei, Xing Che et al.

In this paper we first analyzed the inductive bias underlying the data scattered across complex free energy landscapes (FEL), and exploited it to train deep neural networks which yield reduced and clustered representation for the FEL. Our parametric method, called Information Distilling of Metastability (IDM), is end-to-end differentiable thus scalable to ultra-large dataset. IDM is also a clustering algorithm and is able to cluster the samples in the meantime of reducing the dimensions. Besides, as an unsupervised learning method, IDM differs from many existing dimensionality reduction and clustering methods in that it neither requires a cherry-picked distance metric nor the ground-true number of clusters, and that it can be used to unroll and zoom-in the hierarchical FEL with respect to different timescales. Through multiple experiments, we show that IDM can achieve physically meaningful representations which partition the FEL into well-defined metastable states hence are amenable for downstream tasks such as mechanism analysis and kinetic modeling.